NYTimes:Blogs:: Popular classification
bdanalytics
Date: (Wed) Apr 29, 2015
Introduction:
Data: Source: Training: https://kaggle2.blob.core.windows.net/competitions-data/kaggle/4347/NYTimesBlogTrain.csv New: https://kaggle2.blob.core.windows.net/competitions-data/kaggle/4347/NYTimesBlogTest.csv
Time period:
Synopsis:
Based on analysis utilizing <> techniques,
[](.png)
Potential next steps include:
- Organization:
- Categorize by chunk
- Priority criteria:
- Ease of change
- Impacts report
- Cleans innards
- Bug report
- manage.missing.data chunk:
- cleaner way to manage re-splitting of training vs. new entity
- extract.features chunk:
- Add n-grams for glb_txt_vars
- “RTextTools”, “tau”, “RWeka”, and “textcat” packages
- Convert user-specified mutate code to config specs
- Add n-grams for glb_txt_vars
- fit.models chunk:
- Prediction accuracy scatter graph:
- Add tiles (raw vs. PCA)
- Use shiny for drop-down of “important” features
Use plot.ly for interactive plots ?
- Change .fit suffix of model metrics to .mdl if it’s data independent (e.g. AIC, Adj.R.Squared - is it truly data independent ?, etc.)
- move model_type parameter to myfit_mdl before indep_vars_vctr (keep all model_* together)
- create a custom model for rpart that has minbucket as a tuning parameter
varImp for randomForest crashes in caret version:6.0.41 -> submit bug report
- Probability handling for multinomials vs. desired binomial outcome
- ROCR currently supports only evaluation of binary classification tasks (version 1.0.7)
extensions toward multiclass classification are scheduled for the next release
- Skip trControl.method=“cv” for dummy classifier ?
- Add custom model to caret for a dummy (baseline) classifier (binomial & multinomial) that generates proba/outcomes which mimics the freq distribution of glb_rsp_var values; Right now glb_dmy_glm_mdl always generates most frequent outcome in training data
glm_dmy_mdl should use the same method as glm_sel_mdl until custom dummy classifer is implemented
- fit.all.training chunk:
- myplot_prediction_classification: displays ‘x’ instead of ‘+’ when there are no prediction errors
- Compare glb_sel_mdl vs. glb_fin_mdl:
- varImp
- Prediction differences (shd be minimal ?)
- Move glb_analytics_diag_plots to mydsutils.R: (+) Easier to debug (-) Too many glb vars used
- Add print(ggplot.petrinet(glb_analytics_pn) + coord_flip()) at the end of every major chunk
- Parameterize glb_analytics_pn
- Move glb_impute_missing_data to mydsutils.R: (-) Too many glb vars used; glb_<>_df reassigned
- Replicate myfit_mdl_classification features in myfit_mdl_regression
- Do non-glm methods handle interaction terms ?
- f-score computation for classifiers should be summation across outcomes (not just the desired one ?)
- Add accuracy computation to glb_dmy_mdl in predict.data.new chunk
- Why does splitting fit.data.training.all chunk into separate chunks add an overhead of ~30 secs ? It’s not rbind b/c other chunks have lower elapsed time. Is it the number of plots ?
- Incorporate code chunks in print_sessionInfo
- Test against
- projects in github.com/bdanalytics
- lectures in jhu-datascience track
Analysis:
rm(list=ls())
set.seed(12345)
options(stringsAsFactors=FALSE)
source("~/Dropbox/datascience/R/mydsutils.R")
source("~/Dropbox/datascience/R/myplot.R")
source("~/Dropbox/datascience/R/mypetrinet.R")
# Gather all package requirements here
#suppressPackageStartupMessages(require())
#packageVersion("snow")
#require(sos); findFn("pinv", maxPages=2, sortby="MaxScore")
# Analysis control global variables
glb_trnng_url <- "https://kaggle2.blob.core.windows.net/competitions-data/kaggle/4347/NYTimesBlogTrain.csv"
glb_newdt_url <- "https://kaggle2.blob.core.windows.net/competitions-data/kaggle/4347/NYTimesBlogTest.csv"
glb_is_separate_newent_dataset <- TRUE # or TRUE
glb_split_entity_newent_datasets <- TRUE # or FALSE
glb_split_newdata_method <- "sample" # "condition" or "sample" or "copy"
glb_split_newdata_condition <- "<col_name> <condition_operator> <value>" # or NULL
glb_split_newdata_size_ratio <- 0.3 # > 0 & < 1
glb_split_sample.seed <- 123 # or any integer
glb_max_trnent_obs <- NULL # or any integer
glb_drop_vars <- c(NULL) # or c("<col_name>")
glb_is_regression <- FALSE; glb_is_classification <- TRUE; glb_is_binomial <- TRUE
glb_rsp_var_raw <- "Popular"
# for classification, the response variable has to be a factor
glb_rsp_var <- "Popular.fctr"
# if the response factor is based on numbers e.g (0/1 vs. "A"/"B"),
# caret predict(..., type="prob") crashes
glb_map_rsp_raw_to_var <- function(raw) {
relevel(factor(ifelse(raw == 1, "Y", "N")), as.factor(c("Y", "N")), ref="N")
#as.factor(paste0("B", raw))
#as.factor(raw)
}
glb_map_rsp_raw_to_var(c(1, 1, 0, 0, 0))## [1] Y Y N N N
## Levels: N Yglb_map_rsp_var_to_raw <- function(var) {
as.numeric(var) - 1
#as.numeric(var)
#levels(var)[as.numeric(var)]
#c(" <=50K", " >50K")[as.numeric(var)]
}
glb_map_rsp_var_to_raw(glb_map_rsp_raw_to_var(c(1, 1, 0, 0, 0)))## [1] 1 1 0 0 0if ((glb_rsp_var != glb_rsp_var_raw) & is.null(glb_map_rsp_raw_to_var))
stop("glb_map_rsp_raw_to_var function expected")
glb_rsp_var_out <- paste0(glb_rsp_var, ".predict.") # model_id is appended later
glb_id_vars <- c("UniqueID")
glb_date_vars <- c("PubDate")
glb_is_textual <- TRUE # vs. glb_is_numerical ???
#Sys.setlocale("LC_ALL", "C") # For english
glb_txt_vars <- c("Headline", "Snippet", "Abstract")
glb_append_stop_words <- NULL # or c("<freq_word>")
glb_sprs_threshold <- 0.950 # Ideally, numrows(glb_feats_df) << numrows(glb_trnent_df)
# List transformed vars
glb_exclude_vars_as_features <- c(NULL) # or c("<var_name>")
if (glb_is_textual)
glb_exclude_vars_as_features <- union(glb_exclude_vars_as_features,
glb_txt_vars)
if (glb_rsp_var_raw != glb_rsp_var)
glb_exclude_vars_as_features <- union(glb_exclude_vars_as_features,
glb_rsp_var_raw)
# List feats that shd be excluded due to known causation by prediction variable
glb_exclude_vars_as_features <- union(glb_exclude_vars_as_features,
c(NULL)) # or c("<col_name>")
# List output vars (useful during testing in console)
# glb_exclude_vars_as_features <- union(glb_exclude_vars_as_features,
# grep(glb_rsp_var_out, names(glb_trnent_df), value=TRUE))
glb_impute_na_data <- FALSE # or TRUE
glb_mice_complete.seed <- 144 # or any integer
# rpart: .rnorm messes with the models badly
# caret creates dummy vars for factor feats which messes up the tuning
# - better to feed as.numeric(<feat>.fctr) to caret
# Regression
if (glb_is_regression)
glb_models_method_vctr <- c("lm", "glm", "rpart", "rf") else
# Classification
if (glb_is_binomial)
glb_models_method_vctr <- c("glm", "rpart", "rf") else
glb_models_method_vctr <- c("rpart", "rf")
# Baseline prediction model feature(s)
glb_Baseline_mdl_var <- NULL # or c("<col_name>")
glb_model_metric_terms <- NULL # or matrix(c(
# 0,1,2,3,4,
# 2,0,1,2,3,
# 4,2,0,1,2,
# 6,4,2,0,1,
# 8,6,4,2,0
# ), byrow=TRUE, nrow=5)
glb_model_metric <- NULL # or "<metric_name>"
glb_model_metric_maximize <- NULL # or FALSE (TRUE is not the default for both classification & regression)
glb_model_metric_smmry <- NULL # or function(data, lev=NULL, model=NULL) {
# confusion_mtrx <- t(as.matrix(confusionMatrix(data$pred, data$obs)))
# #print(confusion_mtrx)
# #print(confusion_mtrx * glb_model_metric_terms)
# metric <- sum(confusion_mtrx * glb_model_metric_terms) / nrow(data)
# names(metric) <- glb_model_metric
# return(metric)
# }
glb_tune_models_df <-
rbind(
#data.frame(parameter="cp", min=0.00005, max=0.00005, by=0.000005),
#seq(from=0.01, to=0.01, by=0.01)
#data.frame(parameter="mtry", min=2, max=4, by=1),
data.frame(parameter="dummy", min=2, max=4, by=1)
)
# or NULL
glb_n_cv_folds <- 3 # or NULL
glb_clf_proba_threshold <- NULL # 0.5
# Model selection criteria
if (glb_is_regression)
glb_model_evl_criteria <- c("min.RMSE.OOB", "max.R.sq.OOB", "max.Adj.R.sq.fit")
if (glb_is_classification) {
if (glb_is_binomial)
glb_model_evl_criteria <- c("max.Accuracy.OOB", "max.Kappa.OOB", "min.aic.fit") else
glb_model_evl_criteria <- c("max.Accuracy.OOB", "max.Kappa.OOB")
}
glb_sel_mdl_id <- NULL # or "<model_id_prefix>.<model_method>"
glb_fin_mdl_id <- glb_sel_mdl_id # or "Final"
glb_out_pfx <- "NYTimes_Blogs_Base_"
# Depict process
glb_analytics_pn <- petrinet(name="glb_analytics_pn",
trans_df=data.frame(id=1:6,
name=c("data.training.all","data.new",
"model.selected","model.final",
"data.training.all.prediction","data.new.prediction"),
x=c( -5,-5,-15,-25,-25,-35),
y=c( -5, 5, 0, 0, -5, 5)
),
places_df=data.frame(id=1:4,
name=c("bgn","fit.data.training.all","predict.data.new","end"),
x=c( -0, -20, -30, -40),
y=c( 0, 0, 0, 0),
M0=c( 3, 0, 0, 0)
),
arcs_df=data.frame(
begin=c("bgn","bgn","bgn",
"data.training.all","model.selected","fit.data.training.all",
"fit.data.training.all","model.final",
"data.new","predict.data.new",
"data.training.all.prediction","data.new.prediction"),
end =c("data.training.all","data.new","model.selected",
"fit.data.training.all","fit.data.training.all","model.final",
"data.training.all.prediction","predict.data.new",
"predict.data.new","data.new.prediction",
"end","end")
))
#print(ggplot.petrinet(glb_analytics_pn))
print(ggplot.petrinet(glb_analytics_pn) + coord_flip())## Loading required package: grid
glb_analytics_avl_objs <- NULL
glb_script_tm <- proc.time()
glb_script_df <- data.frame(chunk_label="import_data",
chunk_step_major=1, chunk_step_minor=0,
elapsed=(proc.time() - glb_script_tm)["elapsed"])
print(tail(glb_script_df, 2))## chunk_label chunk_step_major chunk_step_minor elapsed
## elapsed import_data 1 0 0.002Step 1: import data
glb_trnent_df <- myimport_data(url=glb_trnng_url, comment="glb_trnent_df",
force_header=TRUE)## [1] "Reading file ./data/NYTimesBlogTrain.csv..."
## [1] "dimensions of data in ./data/NYTimesBlogTrain.csv: 6,532 rows x 10 cols"
## NewsDesk SectionName SubsectionName
## 1 Business Crosswords/Games
## 2 Culture Arts
## 3 Business Business Day Dealbook
## 4 Business Business Day Dealbook
## 5 Science Health
## 6 Science Health
## Headline
## 1 More School Daze
## 2 New 96-Page Murakami Work Coming in December
## 3 Public Pension Funds Stay Mum on Corporate Expats
## 4 Boot Camp for Bankers
## 5 Of Little Help to Older Knees
## 6 A Benefit of Legal Marijuana
## Snippet
## 1 A puzzle from Ethan Cooper that reminds me that a bill is due.
## 2 The Strange Library will arrive just three and a half months after Mr. Murakamis latest novel, Colorless Tsukuru Tazaki and His Years of Pilgrimage.
## 3 Public pension funds have major stakes in American companies moving overseas to cut their tax bills. But they are saying little about the strategy, which could hurt the nations tax base.
## 4 As they struggle to find new business to bolster sluggish earnings, banks consider the nations 25 million veterans and service members ideal customers.
## 5 Middle-aged and older patients are unlikely to benefit in the long term from surgery to repair tears in the meniscus, pads of cartilage in the knee, a new review of studies has found.
## 6 A new study has found evidence that legal access to marijuana is associated with fewer opioid overdose deaths, but researchers said their findings should not be used as the basis for the wide adoption of legalized cannabis.
## Abstract
## 1 A puzzle from Ethan Cooper that reminds me that a bill is due.
## 2 The Strange Library will arrive just three and a half months after Mr. Murakamis latest novel, Colorless Tsukuru Tazaki and His Years of Pilgrimage.
## 3 Public pension funds have major stakes in American companies moving overseas to cut their tax bills. But they are saying little about the strategy, which could hurt the nations tax base.
## 4 As they struggle to find new business to bolster sluggish earnings, banks consider the nations 25 million veterans and service members ideal customers.
## 5 Middle-aged and older patients are unlikely to benefit in the long term from surgery to repair tears in the meniscus, pads of cartilage in the knee, a new review of studies has found.
## 6 A new study has found evidence that legal access to marijuana is associated with fewer opioid overdose deaths, but researchers said their findings should not be used as the basis for the wide adoption of legalized cannabis.
## WordCount PubDate Popular UniqueID
## 1 508 2014-09-01 22:00:09 1 1
## 2 285 2014-09-01 21:14:07 0 2
## 3 1211 2014-09-01 21:05:36 0 3
## 4 1405 2014-09-01 20:43:34 1 4
## 5 181 2014-09-01 18:58:51 1 5
## 6 245 2014-09-01 18:52:22 1 6
## NewsDesk SectionName SubsectionName
## 226 Styles
## 995
## 2124 TStyle
## 3326 TStyle
## 4752 Business Technology
## 6462 Foreign
## Headline
## 226 For Tavi Gevinson, Fashion Takes a Back Seat, for Now
## 995 Reconsidering What to Call an Extremist Group
## 2124 Paris Fashion Week: Kenzo Spring/Summer 2015
## 3326 The Portable Blue Bottle
## 4752 Monster Moves to Restore a Faded Job Search Brand
## 6462 1889: Priest Questions the Meridian of Greenwich
## Snippet
## 226 Tavi Gevinson, the teenage fashion star turned Broadway actress, wont be much of a player at New York Fashion Week this season.
## 995 Editors have decided to adjust how The Times refer to an Islamic extremist group that controls territory in Syria and Iraq.
## 2124 Scenes from the Paris Fashion Week photo diary of Nina Westervelt.
## 3326 The coffee purveyor has teamed up with its fellow Bay Area-based company Timbuk2 to create a travel kit.
## 4752 Monster, which revolutionized online job hunting in the 1990s, is trying to reinvent itself for the era of Twitter and Facebook with new products that capitalize on social media.
## 6462 From the International Herald Tribune archives: Priest Questions the Meridian of Greenwich in 1889.
## Abstract
## 226 Tavi Gevinson, the teenage fashion star turned Broadway actress, wont be much of a player at New York Fashion Week this season.
## 995 Editors have decided to adjust how The Times refer to an Islamic extremist group that controls territory in Syria and Iraq.
## 2124 Scenes from the Paris Fashion Week photo diary of Nina Westervelt.
## 3326 The coffee purveyor has teamed up with its fellow Bay Area-based company Timbuk2 to create a travel kit.
## 4752 Monster, which revolutionized online job hunting in the 1990s, is trying to reinvent itself for the era of Twitter and Facebook with new products that capitalize on social media.
## 6462 From the International Herald Tribune archives: Priest Questions the Meridian of Greenwich in 1889.
## WordCount PubDate Popular UniqueID
## 226 459 2014-09-04 16:55:57 0 226
## 995 301 2014-09-15 16:05:13 0 995
## 2124 59 2014-09-28 11:20:02 0 2124
## 3326 248 2014-10-14 14:45:55 0 3326
## 4752 995 2014-11-02 07:00:31 0 4752
## 6462 110 2014-11-27 12:00:34 0 6462
## NewsDesk SectionName SubsectionName
## 6527 Foreign
## 6528 Opinion Room For Debate
## 6529 Foreign
## 6530 TStyle
## 6531 Multimedia
## 6532 Business
## Headline
## 6527 1914: Russians Dominate in East Poland
## 6528 Finding a Secretary of Defense
## 6529 1889: Metropolitan Opera House Reopens in New York
## 6530 The Daily Gift: Picasso Plates for Creative Dining
## 6531 Racing From New York to Barcelona
## 6532 Math Anxiety: Why Hollywood Makes Robots of Alan Turing and Other Geniuses
## Snippet
## 6527 From the International Herald Tribune archives: Russians dominate in East Poland in 1914.
## 6528 If Chuck Hagel isn't the right Pentagon chief to respond to an onslaught of global crises, who is?
## 6529 From the International Herald Tribune archives: The Metropolitan Opera House reopens in New York in 1889.
## 6530 Each day until Christmas, the editors of T share a new holiday gift idea.
## 6531 A sailboat race from New York to Barcelona was the setting for a thrilling and sometimes terrifying video about this challenging sport.
## 6532 The visionary who stares at formulas written on walls or mirrors or better yet, thin air has become a Hollywood trope. So has the depiction of the genius who cant connect with real people.
## Abstract
## 6527 From the International Herald Tribune archives: Russians dominate in East Poland in 1914.
## 6528 If Chuck Hagel isn't the right Pentagon chief to respond to an onslaught of global crises, who is?
## 6529 From the International Herald Tribune archives: The Metropolitan Opera House reopens in New York in 1889.
## 6530 Each day until Christmas, the editors of T share a new holiday gift idea.
## 6531 A sailboat race from New York to Barcelona was the setting for a thrilling and sometimes terrifying video about this challenging sport.
## 6532 The visionary who stares at formulas written on walls or mirrors or better yet, thin air has become a Hollywood trope. So has the depiction of the genius who cant connect with real people.
## WordCount PubDate Popular UniqueID
## 6527 176 2014-11-30 13:48:40 0 6527
## 6528 1597 2014-11-30 13:27:23 0 6528
## 6529 214 2014-11-30 09:44:57 0 6529
## 6530 61 2014-11-30 09:00:43 0 6530
## 6531 441 2014-11-30 09:00:22 0 6531
## 6532 921 2014-11-30 07:00:40 0 6532
## 'data.frame': 6532 obs. of 10 variables:
## $ NewsDesk : chr "Business" "Culture" "Business" "Business" ...
## $ SectionName : chr "Crosswords/Games" "Arts" "Business Day" "Business Day" ...
## $ SubsectionName: chr "" "" "Dealbook" "Dealbook" ...
## $ Headline : chr "More School Daze" "New 96-Page Murakami Work Coming in December" "Public Pension Funds Stay Mum on Corporate Expats" "Boot Camp for Bankers" ...
## $ Snippet : chr "A puzzle from Ethan Cooper that reminds me that a bill is due." "The Strange Library will arrive just three and a half months after Mr. Murakamis latest novel, Colorless Tsukuru Tazaki and His"| __truncated__ "Public pension funds have major stakes in American companies moving overseas to cut their tax bills. But they are saying little"| __truncated__ "As they struggle to find new business to bolster sluggish earnings, banks consider the nations 25 million veterans and service "| __truncated__ ...
## $ Abstract : chr "A puzzle from Ethan Cooper that reminds me that a bill is due." "The Strange Library will arrive just three and a half months after Mr. Murakamis latest novel, Colorless Tsukuru Tazaki and His"| __truncated__ "Public pension funds have major stakes in American companies moving overseas to cut their tax bills. But they are saying little"| __truncated__ "As they struggle to find new business to bolster sluggish earnings, banks consider the nations 25 million veterans and service "| __truncated__ ...
## $ WordCount : int 508 285 1211 1405 181 245 258 893 1077 188 ...
## $ PubDate : chr "2014-09-01 22:00:09" "2014-09-01 21:14:07" "2014-09-01 21:05:36" "2014-09-01 20:43:34" ...
## $ Popular : int 1 0 0 1 1 1 0 1 1 0 ...
## $ UniqueID : int 1 2 3 4 5 6 7 8 9 10 ...
## - attr(*, "comment")= chr "glb_trnent_df"
## NULLif (glb_is_separate_newent_dataset) {
glb_newent_df <- myimport_data(url=glb_newdt_url, comment="glb_newent_df",
force_header=TRUE)
# To make plots / stats / checks easier in chunk:inspectORexplore.data
glb_entity_df <- myrbind_df(glb_trnent_df, glb_newent_df);
comment(glb_entity_df) <- "glb_entity_df"
} else {
glb_entity_df <- glb_trnent_df; comment(glb_entity_df) <- "glb_entity_df"
if (!glb_split_entity_newent_datasets) {
stop("Not implemented yet")
glb_newent_df <- glb_trnent_df[sample(1:nrow(glb_trnent_df),
max(2, nrow(glb_trnent_df) / 1000)),]
} else if (glb_split_newdata_method == "condition") {
glb_newent_df <- do.call("subset",
list(glb_trnent_df, parse(text=glb_split_newdata_condition)))
glb_trnent_df <- do.call("subset",
list(glb_trnent_df, parse(text=paste0("!(",
glb_split_newdata_condition,
")"))))
} else if (glb_split_newdata_method == "sample") {
require(caTools)
set.seed(glb_split_sample.seed)
split <- sample.split(glb_trnent_df[, glb_rsp_var_raw],
SplitRatio=(1-glb_split_newdata_size_ratio))
glb_newent_df <- glb_trnent_df[!split, ]
glb_trnent_df <- glb_trnent_df[split ,]
} else if (glb_split_newdata_method == "copy") {
glb_trnent_df <- glb_entity_df
comment(glb_trnent_df) <- "glb_trnent_df"
glb_newent_df <- glb_entity_df
comment(glb_newent_df) <- "glb_newent_df"
} else stop("glb_split_newdata_method should be %in% c('condition', 'sample', 'copy')")
comment(glb_newent_df) <- "glb_newent_df"
myprint_df(glb_newent_df)
str(glb_newent_df)
if (glb_split_entity_newent_datasets) {
myprint_df(glb_trnent_df)
str(glb_trnent_df)
}
} ## [1] "Reading file ./data/NYTimesBlogTest.csv..."
## [1] "dimensions of data in ./data/NYTimesBlogTest.csv: 1,870 rows x 9 cols"
## NewsDesk SectionName SubsectionName
## 1 Culture
## 2 Culture Arts
## 3 Business Crosswords/Games
## 4 Business Business Day Dealbook
## 5 Science Health
## 6 Science Health
## Headline
## 1 'Birdman' Tops the Gothams
## 2 'Sleepy Hollow' Recap: A Not-So-Shocking Death
## 3 Drinking Buddy For Falstaff
## 4 Encouraging Public Service, Through Wall Street's 'Revolving Door'
## 5 Therapy Prevents Repeat Suicide Attempts
## 6 Hoping for a Good Death
## Snippet
## 1 The backstage tale won two awards; Citizenfour, the Edward Snowden documentary, was also a winner.
## 2 In the fall season finale, a question of where the series has many places to go.
## 3 In which Timothy Polin reveals his potty mouth.
## 4 The debate about pay for Wall Street executives who take government jobs appears to be based more on a populist shakedown than on good public policy.
## 5 Short-term psychotherapy may be an effective way to prevent repeated suicide attempts.
## 6 What I hadnt considered before my fathers heart attack was the precise meaning of not wanting to live hooked up to machines.
## Abstract
## 1 The backstage tale won two awards; Citizenfour, the Edward Snowden documentary, was also a winner.
## 2 In the fall season finale, a question of where the series has many places to go.
## 3 In which Timothy Polin reveals his potty mouth.
## 4 The debate about pay for Wall Street executives who take government jobs appears to be based more on a populist shakedown than on good public policy.
## 5 Short-term psychotherapy may be an effective way to prevent repeated suicide attempts.
## 6 What I hadnt considered before my fathers heart attack was the precise meaning of not wanting to live hooked up to machines.
## WordCount PubDate UniqueID
## 1 111 2014-12-01 22:45:24 6533
## 2 558 2014-12-01 22:01:34 6534
## 3 788 2014-12-01 22:00:26 6535
## 4 915 2014-12-01 21:04:13 6536
## 5 213 2014-12-01 19:13:20 6537
## 6 938 2014-12-01 19:05:12 6538
## NewsDesk SectionName SubsectionName
## 3 Business Crosswords/Games
## 725 TStyle
## 731 Business Business Day Dealbook
## 751 TStyle
## 864
## 1376 Business Business Day Small Business
## Headline
## 3 Drinking Buddy For Falstaff
## 725 Ansel Elgort Buttons Up in Brioni
## 731 Didi Dache, a Chinese Ride-Hailing App, Raises $700 Million
## 751 The Daily Gift: A Soft, Colorful Quilt From a Brooklyn Fashion Favorite
## 864 Today in Politics
## 1376 As Health Insurance Evolves, Traditional Brokers Claim They Still Have a Role
## Snippet
## 3 In which Timothy Polin reveals his potty mouth.
## 725 The actor brought a tinge of youthfulness to the classic Italian houses retro-tailored look.
## 731 The Singapore investor Temasek and the Chinese social network operator Tencent are among the leaders of the fund-raising round for a company that says it has 10 times the ridership of Uber.
## 751 Each day until Christmas, the editors of T share a new holiday gift idea.
## 864 The 113th Congress is concluding with partisan brinksmanship and one last mad scramble for votes to pass a $1.1 trillion spending package.
## 1376 Its complex picking insurance for yourself and your family, said a health care policy director for a small-business organization. Its even more complex for a business.
## Abstract
## 3 In which Timothy Polin reveals his potty mouth.
## 725 The actor brought a tinge of youthfulness to the classic Italian houses retro-tailored look.
## 731 The Singapore investor Temasek and the Chinese social network operator Tencent are among the leaders of the fund-raising round for a company that says it has 10 times the ridership of Uber.
## 751 Each day until Christmas, the editors of T share a new holiday gift idea.
## 864 The 113th Congress is concluding with partisan brinksmanship and one last mad scramble for votes to pass a $1.1 trillion spending package.
## 1376 Its complex picking insurance for yourself and your family, said a health care policy director for a small-business organization. Its even more complex for a business.
## WordCount PubDate UniqueID
## 3 788 2014-12-01 22:00:26 6535
## 725 89 2014-12-10 12:30:47 7257
## 731 724 2014-12-10 12:06:32 7263
## 751 85 2014-12-10 09:00:38 7283
## 864 1544 2014-12-11 07:09:25 7396
## 1376 1250 2014-12-18 07:00:05 7908
## NewsDesk SectionName SubsectionName
## 1865
## 1866 Business Technology
## 1867 Metro N.Y. / Region
## 1868 Multimedia
## 1869 Foreign World Asia Pacific
## 1870 Science Health
## Headline
## 1865 Today in Politics
## 1866 Uber Suspends Operations in Spain
## 1867 New York Today: The Year in News
## 1868 New Year, Old Memories, in Times Square
## 1869 Hong Kong Police Criticized After 14-Year-Old's Detention
## 1870 The Super-Short Workout and Other Fitness Trends
## Snippet
## 1865 House Republicans are ending the year on a defensive note over Representative Steve Scalises 2002 speech to a white supremacist group.
## 1866 In a first in the growing pushback against Ubers global expansion, a judges ruling barred telecommunications operators and banks from supporting the companys services.
## 1867 Wednesday: The most read stories of 2014, teeth-chattering cold, and its New Years Eve.
## 1868 What happens when you combine Burning Man, Independence Day fireworks, the last day of school and a full-contact Black Friday sale-a-bration? New Years Eve in Times Square.
## 1869 The authorities have been accused of trying to intimidate young pro-democracy protesters and their families after a 14-year-old girl was detained on suspicion of drawing flowers in chalk near government headquarters and sent to a juvenile home.
## 1870 The big story in exercise science this year was the super-short workout, although many other fitness-related themes emerged in 2014.
## Abstract
## 1865 House Republicans are ending the year on a defensive note over Representative Steve Scalises 2002 speech to a white supremacist group.
## 1866 In a first in the growing pushback against Ubers global expansion, a judges ruling barred telecommunications operators and banks from supporting the companys services.
## 1867 Wednesday: The most read stories of 2014, teeth-chattering cold, and its New Years Eve.
## 1868 What happens when you combine Burning Man, Independence Day fireworks, the last day of school and a full-contact Black Friday sale-a-bration? New Years Eve in Times Square.
## 1869 The authorities have been accused of trying to intimidate young pro-democracy protesters and their families after a 14-year-old girl was detained on suspicion of drawing flowers in chalk near government headquarters and sent to a juvenile home.
## 1870 The big story in exercise science this year was the super-short workout, although many other fitness-related themes emerged in 2014.
## WordCount PubDate UniqueID
## 1865 1616 2014-12-31 07:03:46 8397
## 1866 292 2014-12-31 06:09:32 8398
## 1867 1010 2014-12-31 06:06:58 8399
## 1868 387 2014-12-31 05:00:19 8400
## 1869 717 2014-12-31 04:16:29 8401
## 1870 818 2014-12-31 00:01:10 8402
## 'data.frame': 1870 obs. of 9 variables:
## $ NewsDesk : chr "Culture" "Culture" "Business" "Business" ...
## $ SectionName : chr "" "Arts" "Crosswords/Games" "Business Day" ...
## $ SubsectionName: chr "" "" "" "Dealbook" ...
## $ Headline : chr "'Birdman' Tops the Gothams" "'Sleepy Hollow' Recap: A Not-So-Shocking Death" "Drinking Buddy For Falstaff" "Encouraging Public Service, Through Wall Street's 'Revolving Door'" ...
## $ Snippet : chr "The backstage tale won two awards; Citizenfour, the Edward Snowden documentary, was also a winner." "In the fall season finale, a question of where the series has many places to go." "In which Timothy Polin reveals his potty mouth." "The debate about pay for Wall Street executives who take government jobs appears to be based more on a populist shakedown than "| __truncated__ ...
## $ Abstract : chr "The backstage tale won two awards; Citizenfour, the Edward Snowden documentary, was also a winner." "In the fall season finale, a question of where the series has many places to go." "In which Timothy Polin reveals his potty mouth." "The debate about pay for Wall Street executives who take government jobs appears to be based more on a populist shakedown than "| __truncated__ ...
## $ WordCount : int 111 558 788 915 213 938 1336 2644 752 99 ...
## $ PubDate : chr "2014-12-01 22:45:24" "2014-12-01 22:01:34" "2014-12-01 22:00:26" "2014-12-01 21:04:13" ...
## $ UniqueID : int 6533 6534 6535 6536 6537 6538 6539 6540 6541 6542 ...
## - attr(*, "comment")= chr "glb_newent_df"
## NULLif (!is.null(glb_max_trnent_obs)) {
if (nrow(glb_trnent_df) > glb_max_trnent_obs) {
warning("glb_trnent_df restricted to glb_max_trnent_obs: ",
format(glb_max_trnent_obs, big.mark=","))
org_entity_df <- glb_trnent_df
glb_trnent_df <- org_entity_df[split <-
sample.split(org_entity_df[, glb_rsp_var_raw],
SplitRatio=glb_max_trnent_obs), ]
org_entity_df <- NULL
}
# if (nrow(glb_newent_df) > glb_max_obs) {
# warning("glb_newent_df restricted to glb_max_obs: ", format(glb_max_obs, big.mark=","))
# org_newent_df <- glb_newent_df
# glb_newent_df <- org_newent_df[split <-
# sample.split(org_newent_df[, glb_rsp_var_raw], SplitRatio=glb_max_obs), ]
# org_newent_df <- NULL
# }
}
if (nrow(glb_trnent_df) == nrow(glb_entity_df))
warning("glb_trnent_df same as glb_entity_df")
if (nrow(glb_newent_df) == nrow(glb_entity_df))
warning("glb_newent_df same as glb_entity_df")
if (length(glb_drop_vars) > 0) {
warning("dropping vars: ", paste0(glb_drop_vars, collapse=", "))
glb_entity_df <- glb_entity_df[, setdiff(names(glb_entity_df), glb_drop_vars)]
glb_trnent_df <- glb_trnent_df[, setdiff(names(glb_trnent_df), glb_drop_vars)]
glb_newent_df <- glb_newent_df[, setdiff(names(glb_newent_df), glb_drop_vars)]
}
glb_script_df <- rbind(glb_script_df,
data.frame(chunk_label="cleanse_data",
chunk_step_major=max(glb_script_df$chunk_step_major)+1,
chunk_step_minor=0,
elapsed=(proc.time() - glb_script_tm)["elapsed"]))
print(tail(glb_script_df, 2))## chunk_label chunk_step_major chunk_step_minor elapsed
## elapsed import_data 1 0 0.002
## elapsed1 cleanse_data 2 0 0.864Step 2: cleanse data
glb_script_df <- rbind(glb_script_df,
data.frame(chunk_label="inspectORexplore.data",
chunk_step_major=max(glb_script_df$chunk_step_major),
chunk_step_minor=1,
elapsed=(proc.time() - glb_script_tm)["elapsed"]))
print(tail(glb_script_df, 2))## chunk_label chunk_step_major chunk_step_minor elapsed
## elapsed1 cleanse_data 2 0 0.864
## elapsed2 inspectORexplore.data 2 1 0.903Step 2.1: inspect/explore data
#print(str(glb_trnent_df))
#View(glb_trnent_df)
# List info gathered for various columns
# <col_name>: <description>; <notes>
# Create new features that help diagnostics
# Convert dates to numbers
# typically, dates come in as chars;
# so this must be done before converting chars to factors
myextract_dates_df <- function(df, vars) {
for (var in vars) {
dates_df <- data.frame(.date=strptime(df[, var], "%Y-%m-%d %H:%M:%S"))
dates_df[, paste0(var, ".year")] <- as.numeric(format(dates_df$.date, "%Y"))
dates_df[, paste0(var, ".month")] <- as.numeric(format(dates_df$.date, "%m"))
dates_df[, paste0(var, ".date")] <- as.numeric(format(dates_df$.date, "%d"))
dates_df[, paste0(var, ".wkday")] <- as.numeric(format(dates_df$.date, "%w"))
dates_df[, paste0(var, ".hour")] <- as.numeric(format(dates_df$.date, "%H"))
dates_df[, paste0(var, ".minute")] <- as.numeric(format(dates_df$.date, "%M"))
dates_df[, paste0(var, ".second")] <- as.numeric(format(dates_df$.date, "%S"))
}
#myprint_df(dates_df)
return(subset(dates_df, select=-.date))
}
if (!is.null(glb_date_vars)) {
glb_entity_df <- cbind(glb_entity_df,
myextract_dates_df(glb_entity_df, glb_date_vars))
glb_trnent_df <- cbind(glb_trnent_df,
myextract_dates_df(glb_trnent_df, glb_date_vars))
glb_newent_df <- cbind(glb_newent_df,
myextract_dates_df(glb_newent_df, glb_date_vars))
glb_exclude_vars_as_features <- union(glb_exclude_vars_as_features, glb_date_vars)
}
# Create factors of string variables
str_vars <- sapply(names(glb_trnent_df), function(var)
ifelse(class(glb_trnent_df[, var]) == "character", var, ""))
print(str_vars <- str_vars[str_vars != ""])## NewsDesk SectionName SubsectionName Headline
## "NewsDesk" "SectionName" "SubsectionName" "Headline"
## Snippet Abstract PubDate
## "Snippet" "Abstract" "PubDate"if (length(str_vars <- setdiff(str_vars,
glb_exclude_vars_as_features)) > 0) {
warning("Creating factors of string variables:", paste0(str_vars, collapse=", "))
glb_exclude_vars_as_features <- union(glb_exclude_vars_as_features, str_vars)
for (var in str_vars) {
glb_entity_df[, paste0(var, ".fctr")] <- factor(glb_entity_df[, var],
as.factor(unique(glb_entity_df[, var])))
glb_trnent_df[, paste0(var, ".fctr")] <- factor(glb_trnent_df[, var],
as.factor(unique(glb_entity_df[, var])))
glb_newent_df[, paste0(var, ".fctr")] <- factor(glb_newent_df[, var],
as.factor(unique(glb_entity_df[, var])))
}
}## Warning: Creating factors of string variables:NewsDesk, SectionName,
## SubsectionName# Convert factors to dummy variables
# Build splines require(splines); bsBasis <- bs(training$age, df=3)
add_new_diag_feats <- function(obs_df, ref_df=glb_entity_df) {
require(plyr)
obs_df <- mutate(obs_df,
# <col_name>.NA=is.na(<col_name>),
# <col_name>.fctr=factor(<col_name>,
# as.factor(union(obs_df$<col_name>, obs_twin_df$<col_name>))),
# <col_name>.fctr=relevel(factor(<col_name>,
# as.factor(union(obs_df$<col_name>, obs_twin_df$<col_name>))),
# "<ref_val>"),
# <col2_name>.fctr=relevel(factor(ifelse(<col1_name> == <val>, "<oth_val>", "<ref_val>")),
# as.factor(c("R", "<ref_val>")),
# ref="<ref_val>"),
# This doesn't work - use sapply instead
# <col_name>.fctr_num=grep(<col_name>, levels(<col_name>.fctr)),
#
# Date.my=as.Date(strptime(Date, "%m/%d/%y %H:%M")),
# Year=year(Date.my),
# Month=months(Date.my),
# Weekday=weekdays(Date.my)
# <col_name>.log=log(1 + <col.name>),
# <col_name>=<table>[as.character(<col2_name>)],
# <col_name>=as.numeric(<col2_name>),
.rnorm=rnorm(n=nrow(obs_df))
)
# If levels of a factor are different across obs_df & glb_newent_df; predict.glm fails
# Transformations not handled by mutate
# obs_df$<col_name>.fctr.num <- sapply(1:nrow(obs_df),
# function(row_ix) grep(obs_df[row_ix, "<col_name>"],
# levels(obs_df[row_ix, "<col_name>.fctr"])))
print(summary(obs_df))
print(sapply(names(obs_df), function(col) sum(is.na(obs_df[, col]))))
return(obs_df)
}
glb_entity_df <- add_new_diag_feats(glb_entity_df)## Loading required package: plyr## NewsDesk SectionName SubsectionName
## Length:8402 Length:8402 Length:8402
## Class :character Class :character Class :character
## Mode :character Mode :character Mode :character
##
##
##
##
## Headline Snippet Abstract
## Length:8402 Length:8402 Length:8402
## Class :character Class :character Class :character
## Mode :character Mode :character Mode :character
##
##
##
##
## WordCount PubDate Popular UniqueID
## Min. : 0.0 Length:8402 Min. :0.0000 Min. : 1
## 1st Qu.: 188.0 Class :character 1st Qu.:0.0000 1st Qu.:2101
## Median : 377.0 Mode :character Median :0.0000 Median :4202
## Mean : 528.8 Mean :0.1673 Mean :4202
## 3rd Qu.: 735.0 3rd Qu.:0.0000 3rd Qu.:6302
## Max. :10912.0 Max. :1.0000 Max. :8402
## NA's :1870
## PubDate.year PubDate.month PubDate.date PubDate.wkday
## Min. :2014 Min. : 9.00 Min. : 1.00 Min. :0.000
## 1st Qu.:2014 1st Qu.: 9.00 1st Qu.: 8.00 1st Qu.:2.000
## Median :2014 Median :10.00 Median :15.00 Median :3.000
## Mean :2014 Mean :10.38 Mean :15.17 Mean :2.915
## 3rd Qu.:2014 3rd Qu.:11.00 3rd Qu.:22.00 3rd Qu.:4.000
## Max. :2014 Max. :12.00 Max. :31.00 Max. :6.000
##
## PubDate.hour PubDate.minute PubDate.second NewsDesk.fctr
## Min. : 0.00 Min. : 0.00 Min. : 0.00 :2408
## 1st Qu.: 9.00 1st Qu.: 5.00 1st Qu.:14.00 Business:2026
## Median :12.00 Median :24.00 Median :30.00 Culture : 909
## Mean :12.22 Mean :24.11 Mean :29.49 TStyle : 829
## 3rd Qu.:16.00 3rd Qu.:40.00 3rd Qu.:44.00 OpEd : 680
## Max. :23.00 Max. :59.00 Max. :59.00 Foreign : 477
## (Other) :1073
## SectionName.fctr SubsectionName.fctr .rnorm
## :2899 :6176 Min. :-3.881663
## Business Day:1437 Dealbook :1256 1st Qu.:-0.665043
## Arts : 849 Education : 414 Median :-0.004510
## Opinion : 783 Asia Pacific : 259 Mean :-0.006807
## U.S. : 657 Small Business : 181 3rd Qu.: 0.664125
## Technology : 442 Room For Debate: 82 Max. : 3.356092
## (Other) :1335 (Other) : 34
## NewsDesk SectionName SubsectionName
## 0 0 0
## Headline Snippet Abstract
## 0 0 0
## WordCount PubDate Popular
## 0 0 1870
## UniqueID PubDate.year PubDate.month
## 0 0 0
## PubDate.date PubDate.wkday PubDate.hour
## 0 0 0
## PubDate.minute PubDate.second NewsDesk.fctr
## 0 0 0
## SectionName.fctr SubsectionName.fctr .rnorm
## 0 0 0glb_trnent_df <- add_new_diag_feats(glb_trnent_df)## NewsDesk SectionName SubsectionName
## Length:6532 Length:6532 Length:6532
## Class :character Class :character Class :character
## Mode :character Mode :character Mode :character
##
##
##
##
## Headline Snippet Abstract
## Length:6532 Length:6532 Length:6532
## Class :character Class :character Class :character
## Mode :character Mode :character Mode :character
##
##
##
##
## WordCount PubDate Popular UniqueID
## Min. : 0.0 Length:6532 Min. :0.0000 Min. : 1
## 1st Qu.: 187.0 Class :character 1st Qu.:0.0000 1st Qu.:1634
## Median : 374.0 Mode :character Median :0.0000 Median :3266
## Mean : 524.4 Mean :0.1673 Mean :3266
## 3rd Qu.: 723.2 3rd Qu.:0.0000 3rd Qu.:4899
## Max. :10912.0 Max. :1.0000 Max. :6532
##
## PubDate.year PubDate.month PubDate.date PubDate.wkday
## Min. :2014 Min. : 9.000 Min. : 1.00 Min. :0.000
## 1st Qu.:2014 1st Qu.: 9.000 1st Qu.: 8.00 1st Qu.:2.000
## Median :2014 Median :10.000 Median :16.00 Median :3.000
## Mean :2014 Mean : 9.919 Mean :15.68 Mean :2.931
## 3rd Qu.:2014 3rd Qu.:11.000 3rd Qu.:23.00 3rd Qu.:4.000
## Max. :2014 Max. :11.000 Max. :31.00 Max. :6.000
##
## PubDate.hour PubDate.minute PubDate.second NewsDesk.fctr
## Min. : 0.00 Min. : 0.00 Min. : 0.00 :1846
## 1st Qu.: 9.00 1st Qu.: 6.00 1st Qu.:14.00 Business:1548
## Median :12.00 Median :25.00 Median :29.00 TStyle : 724
## Mean :12.25 Mean :24.42 Mean :29.31 Culture : 676
## 3rd Qu.:16.00 3rd Qu.:40.00 3rd Qu.:44.00 OpEd : 521
## Max. :23.00 Max. :59.00 Max. :59.00 Foreign : 375
## (Other) : 842
## SectionName.fctr SubsectionName.fctr .rnorm
## :2300 :4826 Min. :-3.666157
## Business Day:1092 Dealbook : 952 1st Qu.:-0.655927
## Arts : 675 Education : 325 Median : 0.002704
## Opinion : 607 Asia Pacific : 203 Mean : 0.008495
## U.S. : 505 Small Business : 140 3rd Qu.: 0.664102
## Technology : 330 Room For Debate: 62 Max. : 3.866465
## (Other) :1023 (Other) : 24
## NewsDesk SectionName SubsectionName
## 0 0 0
## Headline Snippet Abstract
## 0 0 0
## WordCount PubDate Popular
## 0 0 0
## UniqueID PubDate.year PubDate.month
## 0 0 0
## PubDate.date PubDate.wkday PubDate.hour
## 0 0 0
## PubDate.minute PubDate.second NewsDesk.fctr
## 0 0 0
## SectionName.fctr SubsectionName.fctr .rnorm
## 0 0 0glb_newent_df <- add_new_diag_feats(glb_newent_df)## NewsDesk SectionName SubsectionName
## Length:1870 Length:1870 Length:1870
## Class :character Class :character Class :character
## Mode :character Mode :character Mode :character
##
##
##
##
## Headline Snippet Abstract WordCount
## Length:1870 Length:1870 Length:1870 Min. : 0.0
## Class :character Class :character Class :character 1st Qu.: 189.0
## Mode :character Mode :character Mode :character Median : 384.5
## Mean : 544.2
## 3rd Qu.: 760.8
## Max. :5956.0
##
## PubDate UniqueID PubDate.year PubDate.month
## Length:1870 Min. :6533 Min. :2014 Min. :12
## Class :character 1st Qu.:7000 1st Qu.:2014 1st Qu.:12
## Mode :character Median :7468 Median :2014 Median :12
## Mean :7468 Mean :2014 Mean :12
## 3rd Qu.:7935 3rd Qu.:2014 3rd Qu.:12
## Max. :8402 Max. :2014 Max. :12
##
## PubDate.date PubDate.wkday PubDate.hour PubDate.minute
## Min. : 1.00 Min. :0.000 Min. : 0.0 Min. : 0.00
## 1st Qu.: 6.00 1st Qu.:2.000 1st Qu.: 9.0 1st Qu.: 4.00
## Median :12.00 Median :3.000 Median :12.0 Median :21.50
## Mean :13.37 Mean :2.858 Mean :12.1 Mean :23.04
## 3rd Qu.:19.00 3rd Qu.:4.000 3rd Qu.:15.0 3rd Qu.:39.00
## Max. :31.00 Max. :6.000 Max. :23.0 Max. :59.00
##
## PubDate.second NewsDesk.fctr SectionName.fctr
## Min. : 0.00 :562 :599
## 1st Qu.:15.00 Business:478 Business Day:345
## Median :30.00 Culture :233 Opinion :176
## Mean :30.14 OpEd :159 Arts :174
## 3rd Qu.:46.00 TStyle :105 U.S. :152
## Max. :59.00 Foreign :102 Technology :112
## (Other) :231 (Other) :312
## SubsectionName.fctr .rnorm
## :1350 Min. :-3.514057
## Dealbook : 304 1st Qu.:-0.705270
## Education : 89 Median :-0.000622
## Asia Pacific : 56 Mean :-0.006081
## Small Business : 41 3rd Qu.: 0.674599
## Room For Debate: 20 Max. : 3.186664
## (Other) : 10
## NewsDesk SectionName SubsectionName
## 0 0 0
## Headline Snippet Abstract
## 0 0 0
## WordCount PubDate UniqueID
## 0 0 0
## PubDate.year PubDate.month PubDate.date
## 0 0 0
## PubDate.wkday PubDate.hour PubDate.minute
## 0 0 0
## PubDate.second NewsDesk.fctr SectionName.fctr
## 0 0 0
## SubsectionName.fctr .rnorm
## 0 0# Histogram of predictor in glb_trnent_df & glb_newent_df
plot_df <- rbind(cbind(glb_trnent_df[, glb_rsp_var_raw, FALSE], data.frame(.data="Training")),
cbind(glb_trnent_df[, glb_rsp_var_raw, FALSE], data.frame(.data="New")))
print(myplot_histogram(plot_df, glb_rsp_var_raw) + facet_wrap(~ .data))## stat_bin: binwidth defaulted to range/30. Use 'binwidth = x' to adjust this.
## stat_bin: binwidth defaulted to range/30. Use 'binwidth = x' to adjust this.
# used later in encode.retype.data chunk
glb_display_class_dstrb <- function(var) {
plot_df <- rbind(cbind(glb_trnent_df[, var, FALSE],
data.frame(.data="Training")),
cbind(glb_trnent_df[, var, FALSE],
data.frame(.data="New")))
xtab_df <- mycreate_xtab(plot_df, c(".data", var))
rownames(xtab_df) <- xtab_df$.data
xtab_df <- subset(xtab_df, select=-.data)
print(xtab_df / rowSums(xtab_df))
}
if (glb_is_classification) glb_display_class_dstrb(glb_rsp_var_raw)## Loading required package: reshape2## Popular.0 Popular.1
## New 0.8326699 0.1673301
## Training 0.8326699 0.1673301# Check for duplicates in glb_id_vars
if (length(glb_id_vars) > 0) {
id_vars_dups_df <- subset(id_vars_df <-
mycreate_tbl_df(glb_entity_df[, glb_id_vars, FALSE], glb_id_vars),
.freq > 1)
} else {
tmp_entity_df <- glb_entity_df
tmp_entity_df$.rownames <- rownames(tmp_entity_df)
id_vars_dups_df <- subset(id_vars_df <-
mycreate_tbl_df(tmp_entity_df[, ".rownames", FALSE], ".rownames"),
.freq > 1)
}
if (nrow(id_vars_dups_df) > 0) {
warning("Duplicates found in glb_id_vars data:", nrow(id_vars_dups_df))
myprint_df(id_vars_dups_df)
} else {
# glb_id_vars are unique across obs in both glb_<>_df
glb_exclude_vars_as_features <- union(glb_exclude_vars_as_features, glb_id_vars)
}
#pairs(subset(glb_trnent_df, select=-c(col_symbol)))
# Check for glb_newent_df & glb_trnent_df features range mismatches
# Other diagnostics:
# print(subset(glb_trnent_df, <col1_name> == max(glb_trnent_df$<col1_name>, na.rm=TRUE) &
# <col2_name> <= mean(glb_trnent_df$<col1_name>, na.rm=TRUE)))
# print(glb_trnent_df[which.max(glb_trnent_df$<col_name>),])
# print(<col_name>_freq_glb_trnent_df <- mycreate_tbl_df(glb_trnent_df, "<col_name>"))
# print(which.min(table(glb_trnent_df$<col_name>)))
# print(which.max(table(glb_trnent_df$<col_name>)))
# print(which.max(table(glb_trnent_df$<col1_name>, glb_trnent_df$<col2_name>)[, 2]))
# print(table(glb_trnent_df$<col1_name>, glb_trnent_df$<col2_name>))
# print(table(is.na(glb_trnent_df$<col1_name>), glb_trnent_df$<col2_name>))
# print(table(sign(glb_trnent_df$<col1_name>), glb_trnent_df$<col2_name>))
# print(mycreate_xtab(glb_trnent_df, <col1_name>))
# print(mycreate_xtab(glb_trnent_df, c(<col1_name>, <col2_name>)))
# print(<col1_name>_<col2_name>_xtab_glb_trnent_df <-
# mycreate_xtab(glb_trnent_df, c("<col1_name>", "<col2_name>")))
# <col1_name>_<col2_name>_xtab_glb_trnent_df[is.na(<col1_name>_<col2_name>_xtab_glb_trnent_df)] <- 0
# print(<col1_name>_<col2_name>_xtab_glb_trnent_df <-
# mutate(<col1_name>_<col2_name>_xtab_glb_trnent_df,
# <col3_name>=(<col1_name> * 1.0) / (<col1_name> + <col2_name>)))
# print(<col2_name>_min_entity_arr <-
# sort(tapply(glb_trnent_df$<col1_name>, glb_trnent_df$<col2_name>, min, na.rm=TRUE)))
# print(<col1_name>_na_by_<col2_name>_arr <-
# sort(tapply(glb_trnent_df$<col1_name>.NA, glb_trnent_df$<col2_name>, mean, na.rm=TRUE)))
# Other plots:
# print(myplot_box(df=glb_trnent_df, ycol_names="<col1_name>"))
# print(myplot_box(df=glb_trnent_df, ycol_names="<col1_name>", xcol_name="<col2_name>"))
# print(myplot_line(subset(glb_trnent_df, Symbol %in% c("KO", "PG")),
# "Date.my", "StockPrice", facet_row_colnames="Symbol") +
# geom_vline(xintercept=as.numeric(as.Date("2003-03-01"))) +
# geom_vline(xintercept=as.numeric(as.Date("1983-01-01")))
# )
# print(myplot_scatter(glb_entity_df, "<col1_name>", "<col2_name>", smooth=TRUE))
# print(myplot_scatter(glb_entity_df, "<col1_name>", "<col2_name>", colorcol_name="<Pred.fctr>") +
# geom_point(data=subset(glb_entity_df, <condition>),
# mapping=aes(x=<x_var>, y=<y_var>), color="red", shape=4, size=5))
glb_script_df <- rbind(glb_script_df,
data.frame(chunk_label="manage_missing_data",
chunk_step_major=max(glb_script_df$chunk_step_major),
chunk_step_minor=glb_script_df[nrow(glb_script_df), "chunk_step_minor"]+1,
elapsed=(proc.time() - glb_script_tm)["elapsed"]))
print(tail(glb_script_df, 2))## chunk_label chunk_step_major chunk_step_minor elapsed
## elapsed2 inspectORexplore.data 2 1 0.903
## elapsed3 manage_missing_data 2 2 2.831Step 2.2: manage missing data
# print(sapply(names(glb_trnent_df), function(col) sum(is.na(glb_trnent_df[, col]))))
# print(sapply(names(glb_newent_df), function(col) sum(is.na(glb_newent_df[, col]))))
# glb_trnent_df <- na.omit(glb_trnent_df)
# glb_newent_df <- na.omit(glb_newent_df)
# df[is.na(df)] <- 0
# Not refactored into mydsutils.R since glb_*_df might be reassigned
glb_impute_missing_data <- function(entity_df, newent_df) {
if (!glb_is_separate_newent_dataset) {
# Combine entity & newent
union_df <- rbind(mutate(entity_df, .src = "entity"),
mutate(newent_df, .src = "newent"))
union_imputed_df <- union_df[, setdiff(setdiff(names(entity_df),
glb_rsp_var),
glb_exclude_vars_as_features)]
print(summary(union_imputed_df))
require(mice)
set.seed(glb_mice_complete.seed)
union_imputed_df <- complete(mice(union_imputed_df))
print(summary(union_imputed_df))
union_df[, names(union_imputed_df)] <- union_imputed_df[, names(union_imputed_df)]
print(summary(union_df))
# union_df$.rownames <- rownames(union_df)
# union_df <- orderBy(~.rownames, union_df)
#
# imp_entity_df <- myimport_data(
# url="<imputed_trnng_url>",
# comment="imp_entity_df", force_header=TRUE, print_diagn=TRUE)
# print(all.equal(subset(union_df, select=-c(.src, .rownames, .rnorm)),
# imp_entity_df))
# Partition again
glb_trnent_df <<- subset(union_df, .src == "entity", select=-c(.src, .rownames))
comment(glb_trnent_df) <- "entity_df"
glb_newent_df <<- subset(union_df, .src == "newent", select=-c(.src, .rownames))
comment(glb_newent_df) <- "newent_df"
# Generate summaries
print(summary(entity_df))
print(sapply(names(entity_df), function(col) sum(is.na(entity_df[, col]))))
print(summary(newent_df))
print(sapply(names(newent_df), function(col) sum(is.na(newent_df[, col]))))
} else stop("Not implemented yet")
}
if (glb_impute_na_data) {
if ((sum(sapply(names(glb_trnent_df),
function(col) sum(is.na(glb_trnent_df[, col])))) > 0) |
(sum(sapply(names(glb_newent_df),
function(col) sum(is.na(glb_newent_df[, col])))) > 0))
glb_impute_missing_data(glb_trnent_df, glb_newent_df)
}
glb_script_df <- rbind(glb_script_df,
data.frame(chunk_label="encodeORretype.data",
chunk_step_major=max(glb_script_df$chunk_step_major),
chunk_step_minor=glb_script_df[nrow(glb_script_df), "chunk_step_minor"]+1,
elapsed=(proc.time() - glb_script_tm)["elapsed"]))
print(tail(glb_script_df, 2))## chunk_label chunk_step_major chunk_step_minor elapsed
## elapsed3 manage_missing_data 2 2 2.831
## elapsed4 encodeORretype.data 2 3 3.123Step 2.3: encode/retype data
# map_<col_name>_df <- myimport_data(
# url="<map_url>",
# comment="map_<col_name>_df", print_diagn=TRUE)
# map_<col_name>_df <- read.csv(paste0(getwd(), "/data/<file_name>.csv"), strip.white=TRUE)
# glb_trnent_df <- mymap_codes(glb_trnent_df, "<from_col_name>", "<to_col_name>",
# map_<to_col_name>_df, map_join_col_name="<map_join_col_name>",
# map_tgt_col_name="<to_col_name>")
# glb_newent_df <- mymap_codes(glb_newent_df, "<from_col_name>", "<to_col_name>",
# map_<to_col_name>_df, map_join_col_name="<map_join_col_name>",
# map_tgt_col_name="<to_col_name>")
# glb_trnent_df$<col_name>.fctr <- factor(glb_trnent_df$<col_name>,
# as.factor(union(glb_trnent_df$<col_name>, glb_newent_df$<col_name>)))
# glb_newent_df$<col_name>.fctr <- factor(glb_newent_df$<col_name>,
# as.factor(union(glb_trnent_df$<col_name>, glb_newent_df$<col_name>)))
if (!is.null(glb_map_rsp_raw_to_var)) {
glb_entity_df[, glb_rsp_var] <-
glb_map_rsp_raw_to_var(glb_entity_df[, glb_rsp_var_raw])
mycheck_map_results(mapd_df=glb_entity_df,
from_col_name=glb_rsp_var_raw, to_col_name=glb_rsp_var)
glb_trnent_df[, glb_rsp_var] <-
glb_map_rsp_raw_to_var(glb_trnent_df[, glb_rsp_var_raw])
mycheck_map_results(mapd_df=glb_trnent_df,
from_col_name=glb_rsp_var_raw, to_col_name=glb_rsp_var)
if (glb_rsp_var %in% names(glb_newent_df)) {
glb_newent_df[, glb_rsp_var] <-
glb_map_rsp_raw_to_var(glb_newent_df[, glb_rsp_var_raw])
mycheck_map_results(mapd_df=glb_newent_df,
from_col_name=glb_rsp_var_raw, to_col_name=glb_rsp_var)
}
if (glb_is_classification) glb_display_class_dstrb(glb_rsp_var)
}## Loading required package: sqldf
## Loading required package: gsubfn
## Loading required package: proto
## Loading required package: RSQLite
## Loading required package: DBI
## Loading required package: tcltk## Popular Popular.fctr .n
## 1 0 N 5439
## 2 NA <NA> 1870
## 3 1 Y 1093## Warning: Removed 1 rows containing missing values (position_stack).
## Popular Popular.fctr .n
## 1 0 N 5439
## 2 1 Y 1093
## Popular.fctr.N Popular.fctr.Y
## New 0.8326699 0.1673301
## Training 0.8326699 0.1673301glb_script_df <- rbind(glb_script_df,
data.frame(chunk_label="extract.features",
chunk_step_major=max(glb_script_df$chunk_step_major)+1,
chunk_step_minor=0,
elapsed=(proc.time() - glb_script_tm)["elapsed"]))
print(tail(glb_script_df, 2))## chunk_label chunk_step_major chunk_step_minor elapsed
## elapsed4 encodeORretype.data 2 3 3.123
## elapsed5 extract.features 3 0 5.551Step 3: extract features
#```{r extract_features, cache=FALSE, eval=glb_is_textual}
# Create new features that help prediction
# <col_name>.lag.2 <- lag(zoo(glb_trnent_df$<col_name>), -2, na.pad=TRUE)
# glb_trnent_df[, "<col_name>.lag.2"] <- coredata(<col_name>.lag.2)
# <col_name>.lag.2 <- lag(zoo(glb_newent_df$<col_name>), -2, na.pad=TRUE)
# glb_newent_df[, "<col_name>.lag.2"] <- coredata(<col_name>.lag.2)
#
# glb_newent_df[1, "<col_name>.lag.2"] <- glb_trnent_df[nrow(glb_trnent_df) - 1,
# "<col_name>"]
# glb_newent_df[2, "<col_name>.lag.2"] <- glb_trnent_df[nrow(glb_trnent_df),
# "<col_name>"]
# glb_entity_df <- mutate(glb_entity_df,
# A.has.http=ifelse(grepl("http",Added,fixed=TRUE), 1, 0)
# )
#
# glb_trnent_df <- mutate(glb_trnent_df,
# )
#
# glb_newent_df <- mutate(glb_newent_df,
# )
if (glb_is_textual) {
require(tm)
glb_corpus_lst <- list(); glb_full_DTM_lst <- list(); glb_sprs_DTM_lst <- list();
for (txt_var in glb_txt_vars) {
print(sprintf("Building corpus for %s...", txt_var))
txt_corpus <- Corpus(VectorSource(glb_entity_df[, txt_var]))
txt_corpus <- tm_map(txt_corpus, tolower)
txt_corpus <- tm_map(txt_corpus, PlainTextDocument)
txt_corpus <- tm_map(txt_corpus, removePunctuation)
txt_corpus <- tm_map(txt_corpus, removeWords,
c(glb_append_stop_words, stopwords("english")))
txt_corpus <- tm_map(txt_corpus, stemDocument)
full_freqs_DTM <- DocumentTermMatrix(txt_corpus)
print(" Full freqs:"); print(full_freqs_DTM)
full_freqs_vctr <- colSums(as.matrix(full_freqs_DTM))
names(full_freqs_vctr) <- dimnames(full_freqs_DTM)[[2]]
full_freqs_df <- as.data.frame(full_freqs_vctr)
names(full_freqs_df) <- "freq.full"
full_freqs_df$term <- rownames(full_freqs_df)
full_freqs_df <- orderBy(~ -freq.full, full_freqs_df)
sprs_freqs_DTM <- removeSparseTerms(full_freqs_DTM, glb_sprs_threshold)
print(" Sparse freqs:"); print(sprs_freqs_DTM)
sprs_freqs_vctr <- colSums(as.matrix(sprs_freqs_DTM))
names(sprs_freqs_vctr) <- dimnames(sprs_freqs_DTM)[[2]]
sprs_freqs_df <- as.data.frame(sprs_freqs_vctr)
names(sprs_freqs_df) <- "freq.sprs"
sprs_freqs_df$term <- rownames(sprs_freqs_df)
sprs_freqs_df <- orderBy(~ -freq.sprs, sprs_freqs_df)
terms_freqs_df <- merge(full_freqs_df, sprs_freqs_df, all.x=TRUE)
melt_freqs_df <- orderBy(~ -value, melt(terms_freqs_df, id.var="term"))
print(ggplot(melt_freqs_df, aes(value, color=variable)) + stat_ecdf() +
geom_hline(yintercept=glb_sprs_threshold, linetype = "dotted"))
print(myplot_hbar(head(melt_freqs_df, 20), "term", "value",
colorcol_name="variable"))
melt_freqs_df <- orderBy(~ -value,
melt(subset(terms_freqs_df, is.na(freq.sprs)), id.var="term"))
print(myplot_hbar(head(melt_freqs_df, 10), "term", "value",
colorcol_name="variable"))
glb_corpus_lst[[txt_var]] <- txt_corpus
glb_full_DTM_lst[[txt_var]] <- full_freqs_DTM
glb_sprs_DTM_lst[[txt_var]] <- sprs_freqs_DTM
}
# Create txt features
if ((length(glb_txt_vars) > 1) &&
(length(unique(pfxs <- sapply(glb_txt_vars,
function(txt) toupper(substr(txt, 1, 1))))) < length(glb_txt_vars)))
stop("Prefixes for corpus freq terms not unique: ", pfxs)
for (txt_var in glb_txt_vars) {
print(sprintf("Binding DTM for %s...", txt_var))
txt_X_df <- as.data.frame(as.matrix(glb_sprs_DTM_lst[[txt_var]]))
colnames(txt_X_df) <- paste(toupper(substr(txt_var, 1, 1)), ".",
make.names(colnames(txt_X_df)), sep="")
rownames(txt_X_df) <- rownames(glb_entity_df) # warning otherwise
glb_entity_df <- cbind(glb_entity_df, txt_X_df)
# Create <txt_var>.has.http
glb_entity_df[, paste(toupper(substr(txt_var, 1, 1)), ".has.http", sep="")] <-
sapply(1:nrow(glb_entity_df),
function(row_ix) ifelse(grepl("http", glb_entity_df[row_ix, txt_var], fixed=TRUE),
1, 0))
# Create <txt_var>.num.chars
glb_entity_df[, paste(toupper(substr(txt_var, 1, 1)), ".num.chars", sep="")] <-
sapply(1:nrow(glb_entity_df),
function(row_ix) nchar(glb_entity_df[row_ix, txt_var]))
# Create <txt_var>.num.words & .num.words.unq
glb_entity_df[, paste(toupper(substr(txt_var, 1, 1)), ".num.words", sep="")] <-
rowSums(as.matrix(glb_full_DTM_lst[[txt_var]]))
glb_entity_df[, paste(toupper(substr(txt_var, 1, 1)), ".num.words.unq", sep="")] <-
rowSums(as.matrix(glb_full_DTM_lst[[txt_var]]) != 0)
for (feat in paste(toupper(substr(txt_var, 1, 1)),
c(".num.chars", ".num.words", ".num.words.unq"), sep="")) {
glb_entity_df[, paste0(feat, ".log")] <- log(1 + glb_entity_df[, feat])
print(myplot_box(glb_entity_df, paste0(feat, ".log"), glb_rsp_var))
}
}
# a working copy of this is reqd in manage.missingdata chunk
union_df <- myrbind_df(mutate(glb_trnent_df, .src = "trnent"),
mutate(glb_newent_df, .src = "newent"))
tmp_entity_df <- glb_entity_df
mrg_id_vars <- ifelse(length(glb_id_vars) > 0, glb_id_vars, ".rownames")
if (mrg_id_vars == ".rownames") {
union_df$.rownames <- rownames(union_df)
tmp_entity_df$.rownames <- rownames(tmp_entity_df)
}
mrg_entity_df <- merge(tmp_entity_df, union_df[, c(".src", mrg_id_vars)])
# Partition again
glb_trnent_df <- subset(mrg_entity_df, .src == "trnent", select=-c(.src))
glb_newent_df <- subset(mrg_entity_df, .src == "newent", select=-c(.src))
if (mrg_id_vars == ".rownames") {
glb_trnent_df <- subset(glb_trnent_df, select=-c(.rownames))
glb_newent_df <- subset(glb_newent_df, select=-c(.rownames))
}
comment(glb_trnent_df) <- "trnent_df"
comment(glb_newent_df) <- "newent_df"
# Generate summaries
# print(summary(glb_entity_df))
# print(sapply(names(glb_entity_df), function(col) sum(is.na(glb_entity_df[, col]))))
# print(summary(glb_trnent_df))
# print(sapply(names(glb_trnent_df), function(col) sum(is.na(glb_trnent_df[, col]))))
# print(summary(glb_newent_df))
# print(sapply(names(glb_newent_df), function(col) sum(is.na(glb_newent_df[, col]))))
}## Loading required package: tm
## Loading required package: NLP
##
## Attaching package: 'NLP'
##
## The following object is masked from 'package:ggplot2':
##
## annotate## [1] "Building corpus for Headline..."
## [1] " Full freqs:"
## <<DocumentTermMatrix (documents: 8402, terms: 9205)>>
## Non-/sparse entries: 44655/77295755
## Sparsity : 100%
## Maximal term length: 31
## Weighting : term frequency (tf)
## [1] " Sparse freqs:"
## <<DocumentTermMatrix (documents: 8402, terms: 1)>>
## Non-/sparse entries: 604/7798
## Sparsity : 93%
## Maximal term length: 3
## Weighting : term frequency (tf)## Warning: Removed 6 rows containing missing values (geom_path).
## [1] "Building corpus for Snippet..."
## [1] " Full freqs:"
## <<DocumentTermMatrix (documents: 8402, terms: 13820)>>
## Non-/sparse entries: 105959/116009681
## Sparsity : 100%
## Maximal term length: 25
## Weighting : term frequency (tf)
## [1] " Sparse freqs:"
## <<DocumentTermMatrix (documents: 8402, terms: 7)>>
## Non-/sparse entries: 4630/54184
## Sparsity : 92%
## Maximal term length: 7
## Weighting : term frequency (tf)## Warning: Removed 6 rows containing missing values (geom_path).
## [1] "Building corpus for Abstract..."
## [1] " Full freqs:"
## <<DocumentTermMatrix (documents: 8402, terms: 13864)>>
## Non-/sparse entries: 106317/116379011
## Sparsity : 100%
## Maximal term length: 112
## Weighting : term frequency (tf)
## [1] " Sparse freqs:"
## <<DocumentTermMatrix (documents: 8402, terms: 7)>>
## Non-/sparse entries: 4639/54175
## Sparsity : 92%
## Maximal term length: 7
## Weighting : term frequency (tf)## Warning: Removed 6 rows containing missing values (geom_path).
## [1] "Binding DTM for Headline..."
## [1] "Binding DTM for Snippet..."
## [1] "Binding DTM for Abstract..."
# print(sapply(names(glb_trnent_df), function(col) sum(is.na(glb_trnent_df[, col]))))
# print(sapply(names(glb_newent_df), function(col) sum(is.na(glb_newent_df[, col]))))
# print(myplot_scatter(glb_trnent_df, "<col1_name>", "<col2_name>", smooth=TRUE))
replay.petrisim(pn=glb_analytics_pn,
replay.trans=(glb_analytics_avl_objs <- c(glb_analytics_avl_objs,
"data.training.all","data.new")), flip_coord=TRUE)## time trans "bgn " "fit.data.training.all " "predict.data.new " "end "
## 0.0000 multiple enabled transitions: data.training.all data.new model.selected firing: data.training.all
## 1.0000 1 2 1 0 0
## 1.0000 multiple enabled transitions: data.training.all data.new model.selected model.final data.training.all.prediction firing: data.new
## 2.0000 2 1 1 1 0
glb_script_df <- rbind(glb_script_df,
data.frame(chunk_label="select.features",
chunk_step_major=max(glb_script_df$chunk_step_major)+1,
chunk_step_minor=0,
elapsed=(proc.time() - glb_script_tm)["elapsed"]))
print(tail(glb_script_df, 2))## chunk_label chunk_step_major chunk_step_minor elapsed
## elapsed5 extract.features 3 0 5.551
## elapsed6 select.features 4 0 85.848Step 4: select features
print(glb_feats_df <- myselect_features(entity_df=glb_trnent_df,
exclude_vars_as_features=glb_exclude_vars_as_features,
rsp_var=glb_rsp_var))## Warning in cor(data.matrix(entity_df[, sel_feats]), y =
## as.numeric(entity_df[, : the standard deviation is zero## id cor.y exclude.as.feat
## Popular Popular 1.000000000 1
## WordCount WordCount 0.257526549 0
## S.num.words.unq.log S.num.words.unq.log -0.244677426 0
## A.num.words.unq.log A.num.words.unq.log -0.244569301 0
## S.num.words.log S.num.words.log -0.239719045 0
## A.num.words.log A.num.words.log -0.239527732 0
## S.num.words.unq S.num.words.unq -0.213378613 0
## A.num.words.unq A.num.words.unq -0.211522561 0
## H.num.words.unq.log H.num.words.unq.log -0.208052027 0
## S.num.words S.num.words -0.207846930 0
## A.num.words A.num.words -0.205674056 0
## A.num.chars.log A.num.chars.log -0.205487403 0
## S.num.chars.log S.num.chars.log -0.205402971 0
## H.num.words.log H.num.words.log -0.204378687 0
## H.num.words.unq H.num.words.unq -0.193599112 0
## H.num.words H.num.words -0.190083728 0
## S.num.chars S.num.chars -0.178433540 0
## A.num.chars A.num.chars -0.176187472 0
## H.num.chars.log H.num.chars.log -0.171062360 0
## PubDate.hour PubDate.hour 0.159167673 0
## H.num.chars H.num.chars -0.147211183 0
## SubsectionName.fctr SubsectionName.fctr -0.137898635 0
## NewsDesk.fctr NewsDesk.fctr -0.115969991 0
## SectionName.fctr SectionName.fctr -0.089486409 0
## S.week S.week -0.084814939 0
## A.week A.week -0.084814939 0
## H.new H.new -0.079839251 0
## A.new A.new -0.069393513 0
## S.new S.new -0.069088180 0
## S.york S.york -0.064751157 0
## A.york A.york -0.064751157 0
## A.will A.will -0.060842964 0
## S.will S.will -0.060392694 0
## A.time A.time -0.057790617 0
## S.time S.time -0.057595102 0
## A.compani A.compani -0.053099633 0
## S.compani S.compani -0.053012962 0
## S.year S.year -0.051235765 0
## A.year A.year -0.051235765 0
## PubDate.wkday PubDate.wkday -0.039801288 0
## PubDate.minute PubDate.minute -0.031469083 0
## PubDate.month PubDate.month 0.019148739 0
## A.has.http A.has.http -0.013592603 0
## PubDate.second PubDate.second -0.012253600 0
## PubDate.date PubDate.date -0.012014100 0
## UniqueID UniqueID 0.011824920 1
## .rnorm .rnorm -0.008703337 0
## PubDate.year PubDate.year NA 0
## H.has.http H.has.http NA 0
## S.has.http S.has.http NA 0
## cor.y.abs
## Popular 1.000000000
## WordCount 0.257526549
## S.num.words.unq.log 0.244677426
## A.num.words.unq.log 0.244569301
## S.num.words.log 0.239719045
## A.num.words.log 0.239527732
## S.num.words.unq 0.213378613
## A.num.words.unq 0.211522561
## H.num.words.unq.log 0.208052027
## S.num.words 0.207846930
## A.num.words 0.205674056
## A.num.chars.log 0.205487403
## S.num.chars.log 0.205402971
## H.num.words.log 0.204378687
## H.num.words.unq 0.193599112
## H.num.words 0.190083728
## S.num.chars 0.178433540
## A.num.chars 0.176187472
## H.num.chars.log 0.171062360
## PubDate.hour 0.159167673
## H.num.chars 0.147211183
## SubsectionName.fctr 0.137898635
## NewsDesk.fctr 0.115969991
## SectionName.fctr 0.089486409
## S.week 0.084814939
## A.week 0.084814939
## H.new 0.079839251
## A.new 0.069393513
## S.new 0.069088180
## S.york 0.064751157
## A.york 0.064751157
## A.will 0.060842964
## S.will 0.060392694
## A.time 0.057790617
## S.time 0.057595102
## A.compani 0.053099633
## S.compani 0.053012962
## S.year 0.051235765
## A.year 0.051235765
## PubDate.wkday 0.039801288
## PubDate.minute 0.031469083
## PubDate.month 0.019148739
## A.has.http 0.013592603
## PubDate.second 0.012253600
## PubDate.date 0.012014100
## UniqueID 0.011824920
## .rnorm 0.008703337
## PubDate.year NA
## H.has.http NA
## S.has.http NAglb_script_df <- rbind(glb_script_df,
data.frame(chunk_label="remove.correlated.features",
chunk_step_major=max(glb_script_df$chunk_step_major),
chunk_step_minor=glb_script_df[nrow(glb_script_df), "chunk_step_minor"]+1,
elapsed=(proc.time() - glb_script_tm)["elapsed"]))
print(tail(glb_script_df, 2))## chunk_label chunk_step_major chunk_step_minor
## elapsed6 select.features 4 0
## elapsed7 remove.correlated.features 4 1
## elapsed
## elapsed6 85.848
## elapsed7 88.177Step 4.1: remove correlated features
print(glb_feats_df <- orderBy(~-cor.y,
myfind_cor_features(feats_df=glb_feats_df, entity_df=glb_trnent_df,
rsp_var=glb_rsp_var,
checkConditionalX=(glb_is_classification && glb_is_binomial))))## Loading required package: caret
## Loading required package: lattice
##
## Attaching package: 'caret'
##
## The following object is masked from 'package:survival':
##
## cluster## [1] "cor(A.week, S.week)=1.0000"
## [1] "cor(Popular.fctr, A.week)=-0.0848"
## [1] "cor(Popular.fctr, S.week)=-0.0848"## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?## Warning in myfind_cor_features(feats_df = glb_feats_df, entity_df =
## glb_trnent_df, : Identified S.week as highly correlated with A.week
## [1] "cor(A.year, S.year)=1.0000"
## [1] "cor(Popular.fctr, A.year)=-0.0512"
## [1] "cor(Popular.fctr, S.year)=-0.0512"## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?## Warning in myfind_cor_features(feats_df = glb_feats_df, entity_df =
## glb_trnent_df, : Identified S.year as highly correlated with A.year
## [1] "cor(A.york, S.york)=1.0000"
## [1] "cor(Popular.fctr, A.york)=-0.0648"
## [1] "cor(Popular.fctr, S.york)=-0.0648"## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?## Warning in myfind_cor_features(feats_df = glb_feats_df, entity_df =
## glb_trnent_df, : Identified S.york as highly correlated with A.york
## [1] "cor(A.time, S.time)=0.9991"
## [1] "cor(Popular.fctr, A.time)=-0.0578"
## [1] "cor(Popular.fctr, S.time)=-0.0576"## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?## Warning in myfind_cor_features(feats_df = glb_feats_df, entity_df =
## glb_trnent_df, : Identified S.time as highly correlated with A.time
## [1] "cor(A.new, S.new)=0.9990"
## [1] "cor(Popular.fctr, A.new)=-0.0694"
## [1] "cor(Popular.fctr, S.new)=-0.0691"## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?## Warning in myfind_cor_features(feats_df = glb_feats_df, entity_df =
## glb_trnent_df, : Identified S.new as highly correlated with A.new
## [1] "cor(A.num.words.unq.log, S.num.words.unq.log)=0.9990"
## [1] "cor(Popular.fctr, A.num.words.unq.log)=-0.2446"
## [1] "cor(Popular.fctr, S.num.words.unq.log)=-0.2447"## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?## Warning in myfind_cor_features(feats_df = glb_feats_df, entity_df =
## glb_trnent_df, : Identified A.num.words.unq.log as highly correlated with
## S.num.words.unq.log
## [1] "cor(A.num.words.log, S.num.words.log)=0.9989"
## [1] "cor(Popular.fctr, A.num.words.log)=-0.2395"
## [1] "cor(Popular.fctr, S.num.words.log)=-0.2397"## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?## Warning in myfind_cor_features(feats_df = glb_feats_df, entity_df =
## glb_trnent_df, : Identified A.num.words.log as highly correlated with
## S.num.words.log
## [1] "cor(A.num.chars.log, S.num.chars.log)=0.9988"
## [1] "cor(Popular.fctr, A.num.chars.log)=-0.2055"
## [1] "cor(Popular.fctr, S.num.chars.log)=-0.2054"## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?## Warning in myfind_cor_features(feats_df = glb_feats_df, entity_df =
## glb_trnent_df, : Identified S.num.chars.log as highly correlated with
## A.num.chars.log
## [1] "cor(A.compani, S.compani)=0.9988"
## [1] "cor(Popular.fctr, A.compani)=-0.0531"
## [1] "cor(Popular.fctr, S.compani)=-0.0530"## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?## Warning in myfind_cor_features(feats_df = glb_feats_df, entity_df =
## glb_trnent_df, : Identified S.compani as highly correlated with A.compani
## [1] "cor(A.will, S.will)=0.9976"
## [1] "cor(Popular.fctr, A.will)=-0.0608"
## [1] "cor(Popular.fctr, S.will)=-0.0604"## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?## Warning in myfind_cor_features(feats_df = glb_feats_df, entity_df =
## glb_trnent_df, : Identified S.will as highly correlated with A.will
## [1] "cor(H.num.words.log, H.num.words.unq.log)=0.9966"
## [1] "cor(Popular.fctr, H.num.words.log)=-0.2044"
## [1] "cor(Popular.fctr, H.num.words.unq.log)=-0.2081"## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?## Warning in myfind_cor_features(feats_df = glb_feats_df, entity_df =
## glb_trnent_df, : Identified H.num.words.log as highly correlated with
## H.num.words.unq.log
## [1] "cor(H.num.words, H.num.words.unq)=0.9963"
## [1] "cor(Popular.fctr, H.num.words)=-0.1901"
## [1] "cor(Popular.fctr, H.num.words.unq)=-0.1936"## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?## Warning in myfind_cor_features(feats_df = glb_feats_df, entity_df =
## glb_trnent_df, : Identified H.num.words as highly correlated with
## H.num.words.unq
## [1] "cor(S.num.words.log, S.num.words.unq.log)=0.9955"
## [1] "cor(Popular.fctr, S.num.words.log)=-0.2397"
## [1] "cor(Popular.fctr, S.num.words.unq.log)=-0.2447"## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?## Warning in myfind_cor_features(feats_df = glb_feats_df, entity_df =
## glb_trnent_df, : Identified S.num.words.log as highly correlated with
## S.num.words.unq.log
## [1] "cor(A.num.words.unq, S.num.words.unq)=0.9944"
## [1] "cor(Popular.fctr, A.num.words.unq)=-0.2115"
## [1] "cor(Popular.fctr, S.num.words.unq)=-0.2134"## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?## Warning in myfind_cor_features(feats_df = glb_feats_df, entity_df =
## glb_trnent_df, : Identified A.num.words.unq as highly correlated with
## S.num.words.unq
## [1] "cor(A.num.words, S.num.words)=0.9936"
## [1] "cor(Popular.fctr, A.num.words)=-0.2057"
## [1] "cor(Popular.fctr, S.num.words)=-0.2078"## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?## Warning in myfind_cor_features(feats_df = glb_feats_df, entity_df =
## glb_trnent_df, : Identified A.num.words as highly correlated with
## S.num.words
## [1] "cor(S.num.words, S.num.words.unq)=0.9920"
## [1] "cor(Popular.fctr, S.num.words)=-0.2078"
## [1] "cor(Popular.fctr, S.num.words.unq)=-0.2134"## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?## Warning in myfind_cor_features(feats_df = glb_feats_df, entity_df =
## glb_trnent_df, : Identified S.num.words as highly correlated with
## S.num.words.unq
## [1] "cor(A.num.chars, S.num.chars)=0.9887"
## [1] "cor(Popular.fctr, A.num.chars)=-0.1762"
## [1] "cor(Popular.fctr, S.num.chars)=-0.1784"## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?## Warning in myfind_cor_features(feats_df = glb_feats_df, entity_df =
## glb_trnent_df, : Identified A.num.chars as highly correlated with
## S.num.chars
## [1] "cor(H.num.words.unq, H.num.words.unq.log)=0.9745"
## [1] "cor(Popular.fctr, H.num.words.unq)=-0.1936"
## [1] "cor(Popular.fctr, H.num.words.unq.log)=-0.2081"## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?## Warning in myfind_cor_features(feats_df = glb_feats_df, entity_df =
## glb_trnent_df, : Identified H.num.words.unq as highly correlated with
## H.num.words.unq.log
## [1] "cor(H.num.chars, H.num.chars.log)=0.9643"
## [1] "cor(Popular.fctr, H.num.chars)=-0.1472"
## [1] "cor(Popular.fctr, H.num.chars.log)=-0.1711"## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?## Warning in myfind_cor_features(feats_df = glb_feats_df, entity_df =
## glb_trnent_df, : Identified H.num.chars as highly correlated with
## H.num.chars.log
## [1] "cor(S.num.words.unq, S.num.words.unq.log)=0.9552"
## [1] "cor(Popular.fctr, S.num.words.unq)=-0.2134"
## [1] "cor(Popular.fctr, S.num.words.unq.log)=-0.2447"## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?## Warning in myfind_cor_features(feats_df = glb_feats_df, entity_df =
## glb_trnent_df, : Identified S.num.words.unq as highly correlated with
## S.num.words.unq.log
## [1] "cor(A.num.chars.log, S.num.words.unq.log)=0.9457"
## [1] "cor(Popular.fctr, A.num.chars.log)=-0.2055"
## [1] "cor(Popular.fctr, S.num.words.unq.log)=-0.2447"## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?## Warning in myfind_cor_features(feats_df = glb_feats_df, entity_df =
## glb_trnent_df, : Identified A.num.chars.log as highly correlated with
## S.num.words.unq.log
## [1] "cor(S.num.chars, S.num.words.unq.log)=0.9043"
## [1] "cor(Popular.fctr, S.num.chars)=-0.1784"
## [1] "cor(Popular.fctr, S.num.words.unq.log)=-0.2447"## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?## Warning in myfind_cor_features(feats_df = glb_feats_df, entity_df =
## glb_trnent_df, : Identified S.num.chars as highly correlated with
## S.num.words.unq.log
## [1] "cor(H.num.chars.log, H.num.words.unq.log)=0.8844"
## [1] "cor(Popular.fctr, H.num.chars.log)=-0.1711"
## [1] "cor(Popular.fctr, H.num.words.unq.log)=-0.2081"## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?
## geom_smooth: Only one unique x value each group.Maybe you want aes(group = 1)?## Warning in myfind_cor_features(feats_df = glb_feats_df, entity_df =
## glb_trnent_df, : Identified H.num.chars.log as highly correlated with
## H.num.words.unq.log
## id cor.y exclude.as.feat
## Popular Popular 1.000000000 1
## WordCount WordCount 0.257526549 0
## PubDate.hour PubDate.hour 0.159167673 0
## PubDate.month PubDate.month 0.019148739 0
## UniqueID UniqueID 0.011824920 1
## .rnorm .rnorm -0.008703337 0
## PubDate.date PubDate.date -0.012014100 0
## PubDate.second PubDate.second -0.012253600 0
## A.has.http A.has.http -0.013592603 0
## PubDate.minute PubDate.minute -0.031469083 0
## PubDate.wkday PubDate.wkday -0.039801288 0
## S.year S.year -0.051235765 0
## A.year A.year -0.051235765 0
## S.compani S.compani -0.053012962 0
## A.compani A.compani -0.053099633 0
## S.time S.time -0.057595102 0
## A.time A.time -0.057790617 0
## S.will S.will -0.060392694 0
## A.will A.will -0.060842964 0
## S.york S.york -0.064751157 0
## A.york A.york -0.064751157 0
## S.new S.new -0.069088180 0
## A.new A.new -0.069393513 0
## H.new H.new -0.079839251 0
## S.week S.week -0.084814939 0
## A.week A.week -0.084814939 0
## SectionName.fctr SectionName.fctr -0.089486409 0
## NewsDesk.fctr NewsDesk.fctr -0.115969991 0
## SubsectionName.fctr SubsectionName.fctr -0.137898635 0
## H.num.chars H.num.chars -0.147211183 0
## H.num.chars.log H.num.chars.log -0.171062360 0
## A.num.chars A.num.chars -0.176187472 0
## S.num.chars S.num.chars -0.178433540 0
## H.num.words H.num.words -0.190083728 0
## H.num.words.unq H.num.words.unq -0.193599112 0
## H.num.words.log H.num.words.log -0.204378687 0
## S.num.chars.log S.num.chars.log -0.205402971 0
## A.num.chars.log A.num.chars.log -0.205487403 0
## A.num.words A.num.words -0.205674056 0
## S.num.words S.num.words -0.207846930 0
## H.num.words.unq.log H.num.words.unq.log -0.208052027 0
## A.num.words.unq A.num.words.unq -0.211522561 0
## S.num.words.unq S.num.words.unq -0.213378613 0
## A.num.words.log A.num.words.log -0.239527732 0
## S.num.words.log S.num.words.log -0.239719045 0
## A.num.words.unq.log A.num.words.unq.log -0.244569301 0
## S.num.words.unq.log S.num.words.unq.log -0.244677426 0
## PubDate.year PubDate.year NA 0
## H.has.http H.has.http NA 0
## S.has.http S.has.http NA 0
## cor.y.abs cor.high.X is.ConditionalX.y
## Popular 1.000000000 <NA> NA
## WordCount 0.257526549 <NA> TRUE
## PubDate.hour 0.159167673 <NA> TRUE
## PubDate.month 0.019148739 <NA> TRUE
## UniqueID 0.011824920 <NA> NA
## .rnorm 0.008703337 <NA> TRUE
## PubDate.date 0.012014100 <NA> TRUE
## PubDate.second 0.012253600 <NA> TRUE
## A.has.http 0.013592603 <NA> FALSE
## PubDate.minute 0.031469083 <NA> TRUE
## PubDate.wkday 0.039801288 <NA> TRUE
## S.year 0.051235765 <NA> TRUE
## A.year 0.051235765 S.year TRUE
## S.compani 0.053012962 <NA> TRUE
## A.compani 0.053099633 S.compani TRUE
## S.time 0.057595102 <NA> TRUE
## A.time 0.057790617 S.time TRUE
## S.will 0.060392694 <NA> TRUE
## A.will 0.060842964 S.will TRUE
## S.york 0.064751157 <NA> TRUE
## A.york 0.064751157 S.york TRUE
## S.new 0.069088180 <NA> TRUE
## A.new 0.069393513 S.new TRUE
## H.new 0.079839251 <NA> TRUE
## S.week 0.084814939 <NA> TRUE
## A.week 0.084814939 S.week TRUE
## SectionName.fctr 0.089486409 <NA> TRUE
## NewsDesk.fctr 0.115969991 <NA> TRUE
## SubsectionName.fctr 0.137898635 <NA> TRUE
## H.num.chars 0.147211183 <NA> TRUE
## H.num.chars.log 0.171062360 H.num.chars TRUE
## A.num.chars 0.176187472 <NA> TRUE
## S.num.chars 0.178433540 A.num.chars TRUE
## H.num.words 0.190083728 <NA> TRUE
## H.num.words.unq 0.193599112 H.num.words TRUE
## H.num.words.log 0.204378687 <NA> TRUE
## S.num.chars.log 0.205402971 <NA> TRUE
## A.num.chars.log 0.205487403 S.num.chars.log TRUE
## A.num.words 0.205674056 <NA> TRUE
## S.num.words 0.207846930 A.num.words TRUE
## H.num.words.unq.log 0.208052027 H.num.chars.log TRUE
## A.num.words.unq 0.211522561 <NA> TRUE
## S.num.words.unq 0.213378613 S.num.words TRUE
## A.num.words.log 0.239527732 <NA> TRUE
## S.num.words.log 0.239719045 A.num.words.log TRUE
## A.num.words.unq.log 0.244569301 <NA> TRUE
## S.num.words.unq.log 0.244677426 S.num.chars TRUE
## PubDate.year NA <NA> FALSE
## H.has.http NA <NA> FALSE
## S.has.http NA <NA> FALSE
## is.cor.y.abs.low
## Popular FALSE
## WordCount FALSE
## PubDate.hour FALSE
## PubDate.month FALSE
## UniqueID FALSE
## .rnorm FALSE
## PubDate.date FALSE
## PubDate.second FALSE
## A.has.http FALSE
## PubDate.minute FALSE
## PubDate.wkday FALSE
## S.year FALSE
## A.year FALSE
## S.compani FALSE
## A.compani FALSE
## S.time FALSE
## A.time FALSE
## S.will FALSE
## A.will FALSE
## S.york FALSE
## A.york FALSE
## S.new FALSE
## A.new FALSE
## H.new FALSE
## S.week FALSE
## A.week FALSE
## SectionName.fctr FALSE
## NewsDesk.fctr FALSE
## SubsectionName.fctr FALSE
## H.num.chars FALSE
## H.num.chars.log FALSE
## A.num.chars FALSE
## S.num.chars FALSE
## H.num.words FALSE
## H.num.words.unq FALSE
## H.num.words.log FALSE
## S.num.chars.log FALSE
## A.num.chars.log FALSE
## A.num.words FALSE
## S.num.words FALSE
## H.num.words.unq.log FALSE
## A.num.words.unq FALSE
## S.num.words.unq FALSE
## A.num.words.log FALSE
## S.num.words.log FALSE
## A.num.words.unq.log FALSE
## S.num.words.unq.log FALSE
## PubDate.year NA
## H.has.http NA
## S.has.http NAglb_script_df <- rbind(glb_script_df,
data.frame(chunk_label="partition.data.training",
chunk_step_major=max(glb_script_df$chunk_step_major)+1,
chunk_step_minor=0,
elapsed=(proc.time() - glb_script_tm)["elapsed"]))
print(tail(glb_script_df, 2))## chunk_label chunk_step_major chunk_step_minor
## elapsed7 remove.correlated.features 4 1
## elapsed8 partition.data.training 5 0
## elapsed
## elapsed7 88.177
## elapsed8 124.117Step 5: partition training data into fit & OOB
if (all(is.na(glb_newent_df[, glb_rsp_var]))) {
require(caTools)
set.seed(glb_split_sample.seed)
split <- sample.split(glb_trnent_df[, glb_rsp_var_raw],
SplitRatio=1 - (nrow(glb_newent_df) * 1.1 / nrow(glb_trnent_df)))
glb_fitent_df <- glb_trnent_df[split, ]
glb_OOBent_df <- glb_trnent_df[!split ,]
} else {
print(sprintf("Newdata contains non-NA data for %s; setting OOB to Newdata",
glb_rsp_var))
glb_fitent_df <- glb_trnent_df; glb_OOBent_df <- glb_newent_df
}## Loading required package: caToolsglb_script_df <- rbind(glb_script_df,
data.frame(chunk_label="fit.models",
chunk_step_major=max(glb_script_df$chunk_step_major)+1,
chunk_step_minor=0,
elapsed=(proc.time() - glb_script_tm)["elapsed"]))
print(tail(glb_script_df, 2))## chunk_label chunk_step_major chunk_step_minor elapsed
## elapsed8 partition.data.training 5 0 124.117
## elapsed9 fit.models 6 0 140.736Step 6: fit models
glb_models_lst <- list(); glb_models_df <- data.frame()
if (glb_is_classification && glb_is_binomial &&
(length(unique(glb_fitent_df[, glb_rsp_var])) < 2))
stop("glb_fitent_df$", glb_rsp_var, ": contains less than 2 unique values: ",
paste0(unique(glb_fitent_df[, glb_rsp_var]), collapse=", "))
max_cor_y_x_var <- orderBy(~ -cor.y.abs,
subset(glb_feats_df, (exclude.as.feat == 0) & !is.cor.y.abs.low))[1, "id"]
if (!is.null(glb_Baseline_mdl_var)) {
if ((max_cor_y_x_var != glb_Baseline_mdl_var) &
(glb_feats_df[max_cor_y_x_var, "cor.y.abs"] >
glb_feats_df[glb_Baseline_mdl_var, "cor.y.abs"]))
stop(max_cor_y_x_var, " has a lower correlation with ", glb_rsp_var,
" than the Baseline var: ", glb_Baseline_mdl_var)
}
glb_model_type <- ifelse(glb_is_regression, "regression", "classification")
# Baseline
if (!is.null(glb_Baseline_mdl_var))
ret_lst <- myfit_mdl_fn(model_id="Baseline", model_method="mybaseln_classfr",
indep_vars_vctr=glb_Baseline_mdl_var,
rsp_var=glb_rsp_var, rsp_var_out=glb_rsp_var_out,
fit_df=glb_fitent_df, OOB_df=glb_OOBent_df)
# Most Frequent Outcome "MFO" model: mean(y) for regression
# Not using caret's nullModel since model stats not avl
# Cannot use rpart for multinomial classification since it predicts non-MFO
ret_lst <- myfit_mdl(model_id="MFO",
model_method=ifelse(glb_is_regression, "lm", "myMFO_classfr"),
model_type=glb_model_type,
indep_vars_vctr=".rnorm",
rsp_var=glb_rsp_var, rsp_var_out=glb_rsp_var_out,
fit_df=glb_fitent_df, OOB_df=glb_OOBent_df)## [1] "fitting model: MFO.myMFO_classfr"
## [1] " indep_vars: .rnorm"
## Fitting parameter = none on full training set
## [1] "in MFO.Classifier$fit"
## [1] "unique.vals:"
## [1] N Y
## Levels: N Y
## [1] "unique.prob:"
## y
## N Y
## 0.8326257 0.1673743
## [1] "MFO.val:"
## [1] "N"
## Length Class Mode
## unique.vals 2 factor numeric
## unique.prob 2 -none- numeric
## MFO.val 1 -none- character
## x.names 1 -none- character
## xNames 1 -none- character
## problemType 1 -none- character
## tuneValue 1 data.frame list
## obsLevels 2 -none- character
## [1] " calling mypredict_mdl for fit:"## Loading required package: ROCR
## Loading required package: gplots
##
## Attaching package: 'gplots'
##
## The following object is masked from 'package:stats':
##
## lowess## [1] "entr MFO.Classifier$predict"
## [1] "exit MFO.Classifier$predict"
## [1] "in MFO.Classifier$prob"
## N Y
## 1 0.8326257 0.1673743
## 2 0.8326257 0.1673743
## 3 0.8326257 0.1673743
## 4 0.8326257 0.1673743
## 5 0.8326257 0.1673743
## 6 0.8326257 0.1673743
## [1] "Classifier Probability Threshold: 0.5000 to maximize f.score.fit"
## Popular.fctr Popular.fctr.predict.MFO.myMFO_classfr.N
## 1 N 3726
## 2 Y 749
## Reference
## Prediction N Y
## N 3726 749
## Y 0 0
## Popular.fctr Popular.fctr.predict.MFO.myMFO_classfr.N
## 1 N 3726
## 2 Y 749
## Popular.fctr.predict.MFO.myMFO_classfr.Y
## 1 0
## 2 0
## [1] "in mypredict_mdl: calling confusionMatrix..."
## [1] " unique(df[, rsp_var_out]: "
## [1] N
## Levels: N Y
## [1] " unique(df[, rsp_var ]: "
## [1] Y N
## Levels: N Y
## Prediction
## Reference N Y
## N 3726 0
## Y 749 0
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 8.326257e-01 0.000000e+00 8.213602e-01 8.434553e-01 8.326257e-01
## AccuracyPValue McnemarPValue
## 5.097571e-01 1.800616e-164
## [1] " calling mypredict_mdl for OOB:"
## [1] "entr MFO.Classifier$predict"
## [1] "exit MFO.Classifier$predict"
## [1] "in MFO.Classifier$prob"
## N Y
## 1 0.8326257 0.1673743
## 2 0.8326257 0.1673743
## 3 0.8326257 0.1673743
## 4 0.8326257 0.1673743
## 5 0.8326257 0.1673743
## 6 0.8326257 0.1673743
## [1] "Classifier Probability Threshold: 0.5000 to maximize f.score.OOB"
## Popular.fctr Popular.fctr.predict.MFO.myMFO_classfr.N
## 1 N 1713
## 2 Y 344
## Reference
## Prediction N Y
## N 1713 344
## Y 0 0
## Popular.fctr Popular.fctr.predict.MFO.myMFO_classfr.N
## 1 N 1713
## 2 Y 344
## Popular.fctr.predict.MFO.myMFO_classfr.Y
## 1 0
## 2 0
## [1] "in mypredict_mdl: calling confusionMatrix..."
## [1] " unique(df[, rsp_var_out]: "
## [1] N
## Levels: N Y
## [1] " unique(df[, rsp_var ]: "
## [1] N Y
## Levels: N Y
## Prediction
## Reference N Y
## N 1713 0
## Y 344 0
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 8.327662e-01 0.000000e+00 8.159247e-01 8.486533e-01 8.327662e-01
## AccuracyPValue McnemarPValue
## 5.143944e-01 2.337097e-76
## model_id model_method feats max.nTuningRuns
## 1 MFO.myMFO_classfr myMFO_classfr .rnorm 0
## min.elapsedtime.everything min.elapsedtime.final max.auc.fit
## 1 0.46 0.003 0.5
## opt.prob.threshold.fit max.f.score.fit max.Accuracy.fit
## 1 0.5 0 0.8326257
## max.AccuracyLower.fit max.AccuracyUpper.fit max.Kappa.fit max.auc.OOB
## 1 0.8213602 0.8434553 0 0.5
## opt.prob.threshold.OOB max.f.score.OOB max.Accuracy.OOB
## 1 0.5 0 0.8327662
## max.AccuracyLower.OOB max.AccuracyUpper.OOB max.Kappa.OOB
## 1 0.8159247 0.8486533 0if (glb_is_classification)
# "random" model - only for classification;
# none needed for regression since it is same as MFO
ret_lst <- myfit_mdl(model_id="Random", model_method="myrandom_classfr",
model_type=glb_model_type,
indep_vars_vctr=".rnorm",
rsp_var=glb_rsp_var, rsp_var_out=glb_rsp_var_out,
fit_df=glb_fitent_df, OOB_df=glb_OOBent_df)## [1] "fitting model: Random.myrandom_classfr"
## [1] " indep_vars: .rnorm"
## Fitting parameter = none on full training set
## Length Class Mode
## unique.vals 2 factor numeric
## unique.prob 2 table numeric
## xNames 1 -none- character
## problemType 1 -none- character
## tuneValue 1 data.frame list
## obsLevels 2 -none- character
## [1] " calling mypredict_mdl for fit:"
## [1] "in Random.Classifier$prob"
## Reference
## Prediction N Y
## N 0 0
## Y 3726 749
## Popular.fctr Popular.fctr.predict.Random.myrandom_classfr.N
## 1 N 0
## 2 Y 0
## Popular.fctr.predict.Random.myrandom_classfr.Y
## 1 3726
## 2 749
## Reference
## Prediction N Y
## N 0 0
## Y 3726 749
## Popular.fctr Popular.fctr.predict.Random.myrandom_classfr.N
## 1 N 0
## 2 Y 0
## Popular.fctr.predict.Random.myrandom_classfr.Y
## 1 3726
## 2 749
## Reference
## Prediction N Y
## N 3061 619
## Y 665 130
## Popular.fctr Popular.fctr.predict.Random.myrandom_classfr.N
## 1 N 3061
## 2 Y 619
## Popular.fctr.predict.Random.myrandom_classfr.Y
## 1 665
## 2 130
## Reference
## Prediction N Y
## N 3061 619
## Y 665 130
## Popular.fctr Popular.fctr.predict.Random.myrandom_classfr.N
## 1 N 3061
## 2 Y 619
## Popular.fctr.predict.Random.myrandom_classfr.Y
## 1 665
## 2 130
## Reference
## Prediction N Y
## N 3061 619
## Y 665 130
## Popular.fctr Popular.fctr.predict.Random.myrandom_classfr.N
## 1 N 3061
## 2 Y 619
## Popular.fctr.predict.Random.myrandom_classfr.Y
## 1 665
## 2 130
## Reference
## Prediction N Y
## N 3061 619
## Y 665 130
## Popular.fctr Popular.fctr.predict.Random.myrandom_classfr.N
## 1 N 3061
## 2 Y 619
## Popular.fctr.predict.Random.myrandom_classfr.Y
## 1 665
## 2 130
## Reference
## Prediction N Y
## N 3061 619
## Y 665 130
## Popular.fctr Popular.fctr.predict.Random.myrandom_classfr.N
## 1 N 3061
## 2 Y 619
## Popular.fctr.predict.Random.myrandom_classfr.Y
## 1 665
## 2 130
## Reference
## Prediction N Y
## N 3061 619
## Y 665 130
## Popular.fctr Popular.fctr.predict.Random.myrandom_classfr.N
## 1 N 3061
## 2 Y 619
## Popular.fctr.predict.Random.myrandom_classfr.Y
## 1 665
## 2 130
## Reference
## Prediction N Y
## N 3061 619
## Y 665 130
## Popular.fctr Popular.fctr.predict.Random.myrandom_classfr.N
## 1 N 3061
## 2 Y 619
## Popular.fctr.predict.Random.myrandom_classfr.Y
## 1 665
## 2 130
## Reference
## Prediction N Y
## N 3726 749
## Y 0 0
## Popular.fctr Popular.fctr.predict.Random.myrandom_classfr.N
## 1 N 3726
## 2 Y 749
## Popular.fctr.predict.Random.myrandom_classfr.Y
## 1 0
## 2 0
## Reference
## Prediction N Y
## N 3726 749
## Y 0 0
## Popular.fctr Popular.fctr.predict.Random.myrandom_classfr.N
## 1 N 3726
## 2 Y 749
## Popular.fctr.predict.Random.myrandom_classfr.Y
## 1 0
## 2 0
## threshold f.score
## 1 0.0 0.2867534
## 2 0.1 0.2867534
## 3 0.2 0.1683938
## 4 0.3 0.1683938
## 5 0.4 0.1683938
## 6 0.5 0.1683938
## 7 0.6 0.1683938
## 8 0.7 0.1683938
## 9 0.8 0.1683938
## 10 0.9 0.0000000
## 11 1.0 0.0000000
## [1] "Classifier Probability Threshold: 0.1000 to maximize f.score.fit"
## Popular.fctr Popular.fctr.predict.Random.myrandom_classfr.Y
## 1 N 3726
## 2 Y 749
## Reference
## Prediction N Y
## N 0 0
## Y 3726 749
## Popular.fctr Popular.fctr.predict.Random.myrandom_classfr.N
## 1 N 0
## 2 Y 0
## Popular.fctr.predict.Random.myrandom_classfr.Y
## 1 3726
## 2 749
## [1] "in mypredict_mdl: calling confusionMatrix..."
## [1] " unique(df[, rsp_var_out]: "
## [1] Y
## Levels: N Y
## [1] " unique(df[, rsp_var ]: "
## [1] Y N
## Levels: N Y
## Prediction
## Reference N Y
## N 0 3726
## Y 0 749
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 0.1673743 0.0000000 0.1565447 0.1786398 0.8326257
## AccuracyPValue McnemarPValue
## 1.0000000 0.0000000
## [1] " calling mypredict_mdl for OOB:"
## [1] "in Random.Classifier$prob"
## Reference
## Prediction N Y
## N 0 0
## Y 1713 344
## Popular.fctr Popular.fctr.predict.Random.myrandom_classfr.N
## 1 N 0
## 2 Y 0
## Popular.fctr.predict.Random.myrandom_classfr.Y
## 1 1713
## 2 344
## Reference
## Prediction N Y
## N 0 0
## Y 1713 344
## Popular.fctr Popular.fctr.predict.Random.myrandom_classfr.N
## 1 N 0
## 2 Y 0
## Popular.fctr.predict.Random.myrandom_classfr.Y
## 1 1713
## 2 344
## Reference
## Prediction N Y
## N 1408 295
## Y 305 49
## Popular.fctr Popular.fctr.predict.Random.myrandom_classfr.N
## 1 N 1408
## 2 Y 295
## Popular.fctr.predict.Random.myrandom_classfr.Y
## 1 305
## 2 49
## Reference
## Prediction N Y
## N 1408 295
## Y 305 49
## Popular.fctr Popular.fctr.predict.Random.myrandom_classfr.N
## 1 N 1408
## 2 Y 295
## Popular.fctr.predict.Random.myrandom_classfr.Y
## 1 305
## 2 49
## Reference
## Prediction N Y
## N 1408 295
## Y 305 49
## Popular.fctr Popular.fctr.predict.Random.myrandom_classfr.N
## 1 N 1408
## 2 Y 295
## Popular.fctr.predict.Random.myrandom_classfr.Y
## 1 305
## 2 49
## Reference
## Prediction N Y
## N 1408 295
## Y 305 49
## Popular.fctr Popular.fctr.predict.Random.myrandom_classfr.N
## 1 N 1408
## 2 Y 295
## Popular.fctr.predict.Random.myrandom_classfr.Y
## 1 305
## 2 49
## Reference
## Prediction N Y
## N 1408 295
## Y 305 49
## Popular.fctr Popular.fctr.predict.Random.myrandom_classfr.N
## 1 N 1408
## 2 Y 295
## Popular.fctr.predict.Random.myrandom_classfr.Y
## 1 305
## 2 49
## Reference
## Prediction N Y
## N 1408 295
## Y 305 49
## Popular.fctr Popular.fctr.predict.Random.myrandom_classfr.N
## 1 N 1408
## 2 Y 295
## Popular.fctr.predict.Random.myrandom_classfr.Y
## 1 305
## 2 49
## Reference
## Prediction N Y
## N 1408 295
## Y 305 49
## Popular.fctr Popular.fctr.predict.Random.myrandom_classfr.N
## 1 N 1408
## 2 Y 295
## Popular.fctr.predict.Random.myrandom_classfr.Y
## 1 305
## 2 49
## Reference
## Prediction N Y
## N 1713 344
## Y 0 0
## Popular.fctr Popular.fctr.predict.Random.myrandom_classfr.N
## 1 N 1713
## 2 Y 344
## Popular.fctr.predict.Random.myrandom_classfr.Y
## 1 0
## 2 0
## Reference
## Prediction N Y
## N 1713 344
## Y 0 0
## Popular.fctr Popular.fctr.predict.Random.myrandom_classfr.N
## 1 N 1713
## 2 Y 344
## Popular.fctr.predict.Random.myrandom_classfr.Y
## 1 0
## 2 0
## threshold f.score
## 1 0.0 0.2865473
## 2 0.1 0.2865473
## 3 0.2 0.1404011
## 4 0.3 0.1404011
## 5 0.4 0.1404011
## 6 0.5 0.1404011
## 7 0.6 0.1404011
## 8 0.7 0.1404011
## 9 0.8 0.1404011
## 10 0.9 0.0000000
## 11 1.0 0.0000000
## [1] "Classifier Probability Threshold: 0.1000 to maximize f.score.OOB"
## Popular.fctr Popular.fctr.predict.Random.myrandom_classfr.Y
## 1 N 1713
## 2 Y 344
## Reference
## Prediction N Y
## N 0 0
## Y 1713 344
## Popular.fctr Popular.fctr.predict.Random.myrandom_classfr.N
## 1 N 0
## 2 Y 0
## Popular.fctr.predict.Random.myrandom_classfr.Y
## 1 1713
## 2 344
## [1] "in mypredict_mdl: calling confusionMatrix..."
## [1] " unique(df[, rsp_var_out]: "
## [1] Y
## Levels: N Y
## [1] " unique(df[, rsp_var ]: "
## [1] N Y
## Levels: N Y
## Prediction
## Reference N Y
## N 0 1713
## Y 0 344
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 0.1672338 0.0000000 0.1513467 0.1840753 0.8327662
## AccuracyPValue McnemarPValue
## 1.0000000 0.0000000
## model_id model_method feats max.nTuningRuns
## 1 Random.myrandom_classfr myrandom_classfr .rnorm 0
## min.elapsedtime.everything min.elapsedtime.final max.auc.fit
## 1 0.336 0.002 0.4975446
## opt.prob.threshold.fit max.f.score.fit max.Accuracy.fit
## 1 0.1 0.2867534 0.1673743
## max.AccuracyLower.fit max.AccuracyUpper.fit max.Kappa.fit max.auc.OOB
## 1 0.1565447 0.1786398 0 0.4821958
## opt.prob.threshold.OOB max.f.score.OOB max.Accuracy.OOB
## 1 0.1 0.2865473 0.1672338
## max.AccuracyLower.OOB max.AccuracyUpper.OOB max.Kappa.OOB
## 1 0.1513467 0.1840753 0# Any models that have tuning parameters has "better" results with cross-validation
# (except rf) & "different" results for different outcome metrics
# Max.cor.Y
# Check impact of cv
# rpart is not a good candidate since caret does not optimize cp (only tuning parameter of rpart) well
ret_lst <- myfit_mdl(model_id="Max.cor.Y.cv.0",
model_method="rpart",
model_type=glb_model_type,
indep_vars_vctr=max_cor_y_x_var,
rsp_var=glb_rsp_var, rsp_var_out=glb_rsp_var_out,
fit_df=glb_fitent_df, OOB_df=glb_OOBent_df)## [1] "fitting model: Max.cor.Y.cv.0.rpart"
## [1] " indep_vars: WordCount"## Loading required package: rpart## Fitting cp = 0.00223 on full training set## Loading required package: rpart.plot
## Call:
## rpart(formula = .outcome ~ ., control = list(minsplit = 20, minbucket = 7,
## cp = 0, maxcompete = 4, maxsurrogate = 5, usesurrogate = 2,
## surrogatestyle = 0, maxdepth = 30, xval = 0))
## n= 4475
##
## CP nsplit rel error
## 1 0.002225189 0 1
##
## Node number 1: 4475 observations
## predicted class=N expected loss=0.1673743 P(node) =1
## class counts: 3726 749
## probabilities: 0.833 0.167
##
## n= 4475
##
## node), split, n, loss, yval, (yprob)
## * denotes terminal node
##
## 1) root 4475 749 N (0.8326257 0.1673743) *
## [1] " calling mypredict_mdl for fit:"
## [1] "Classifier Probability Threshold: 0.5000 to maximize f.score.fit"
## Popular.fctr Popular.fctr.predict.Max.cor.Y.cv.0.rpart.N
## 1 N 3726
## 2 Y 749
## Reference
## Prediction N Y
## N 3726 749
## Y 0 0
## Popular.fctr Popular.fctr.predict.Max.cor.Y.cv.0.rpart.N
## 1 N 3726
## 2 Y 749
## Popular.fctr.predict.Max.cor.Y.cv.0.rpart.Y
## 1 0
## 2 0
## [1] "in mypredict_mdl: calling confusionMatrix..."
## [1] " unique(df[, rsp_var_out]: "
## [1] N
## Levels: N Y
## [1] " unique(df[, rsp_var ]: "
## [1] Y N
## Levels: N Y
## Prediction
## Reference N Y
## N 3726 0
## Y 749 0
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 8.326257e-01 0.000000e+00 8.213602e-01 8.434553e-01 8.326257e-01
## AccuracyPValue McnemarPValue
## 5.097571e-01 1.800616e-164
## [1] " calling mypredict_mdl for OOB:"
## [1] "Classifier Probability Threshold: 0.5000 to maximize f.score.OOB"
## Popular.fctr Popular.fctr.predict.Max.cor.Y.cv.0.rpart.N
## 1 N 1713
## 2 Y 344
## Reference
## Prediction N Y
## N 1713 344
## Y 0 0
## Popular.fctr Popular.fctr.predict.Max.cor.Y.cv.0.rpart.N
## 1 N 1713
## 2 Y 344
## Popular.fctr.predict.Max.cor.Y.cv.0.rpart.Y
## 1 0
## 2 0
## [1] "in mypredict_mdl: calling confusionMatrix..."
## [1] " unique(df[, rsp_var_out]: "
## [1] N
## Levels: N Y
## [1] " unique(df[, rsp_var ]: "
## [1] N Y
## Levels: N Y
## Prediction
## Reference N Y
## N 1713 0
## Y 344 0
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 8.327662e-01 0.000000e+00 8.159247e-01 8.486533e-01 8.327662e-01
## AccuracyPValue McnemarPValue
## 5.143944e-01 2.337097e-76
## model_id model_method feats max.nTuningRuns
## 1 Max.cor.Y.cv.0.rpart rpart WordCount 0
## min.elapsedtime.everything min.elapsedtime.final max.auc.fit
## 1 0.753 0.063 0.5
## opt.prob.threshold.fit max.f.score.fit max.Accuracy.fit
## 1 0.5 0 0.8326257
## max.AccuracyLower.fit max.AccuracyUpper.fit max.Kappa.fit max.auc.OOB
## 1 0.8213602 0.8434553 0 0.5
## opt.prob.threshold.OOB max.f.score.OOB max.Accuracy.OOB
## 1 0.5 0 0.8327662
## max.AccuracyLower.OOB max.AccuracyUpper.OOB max.Kappa.OOB
## 1 0.8159247 0.8486533 0ret_lst <- myfit_mdl(model_id="Max.cor.Y.cv.0.cp.0",
model_method="rpart",
model_type=glb_model_type,
indep_vars_vctr=max_cor_y_x_var,
rsp_var=glb_rsp_var, rsp_var_out=glb_rsp_var_out,
fit_df=glb_fitent_df, OOB_df=glb_OOBent_df,
n_cv_folds=0,
tune_models_df=data.frame(parameter="cp", min=0.0, max=0.0, by=0.1))## [1] "fitting model: Max.cor.Y.cv.0.cp.0.rpart"
## [1] " indep_vars: WordCount"
## Fitting cp = 0 on full training set
## Call:
## rpart(formula = .outcome ~ ., control = list(minsplit = 20, minbucket = 7,
## cp = 0, maxcompete = 4, maxsurrogate = 5, usesurrogate = 2,
## surrogatestyle = 0, maxdepth = 30, xval = 0))
## n= 4475
##
## CP nsplit rel error
## 1 0.0022251891 0 1.0000000
## 2 0.0020026702 17 0.9559413
## 3 0.0019073050 19 0.9519359
## 4 0.0017801513 33 0.9238985
## 5 0.0013351135 39 0.9132176
## 6 0.0008900757 49 0.8998665
## 7 0.0002670227 54 0.8945260
## 8 0.0000000000 59 0.8931909
##
## Variable importance
## WordCount
## 100
##
## Node number 1: 4475 observations, complexity param=0.002225189
## predicted class=N expected loss=0.1673743 P(node) =1
## class counts: 3726 749
## probabilities: 0.833 0.167
## left son=2 (3291 obs) right son=3 (1184 obs)
## Primary splits:
## WordCount < 683.5 to the left, improve=113.0263, (0 missing)
##
## Node number 2: 3291 observations
## predicted class=N expected loss=0.09996961 P(node) =0.735419
## class counts: 2962 329
## probabilities: 0.900 0.100
##
## Node number 3: 1184 observations, complexity param=0.002225189
## predicted class=N expected loss=0.3547297 P(node) =0.264581
## class counts: 764 420
## probabilities: 0.645 0.355
## left son=6 (190 obs) right son=7 (994 obs)
## Primary splits:
## WordCount < 782.5 to the left, improve=2.973082, (0 missing)
##
## Node number 6: 190 observations, complexity param=0.001780151
## predicted class=N expected loss=0.2736842 P(node) =0.0424581
## class counts: 138 52
## probabilities: 0.726 0.274
## left son=12 (73 obs) right son=13 (117 obs)
## Primary splits:
## WordCount < 749.5 to the right, improve=2.166979, (0 missing)
##
## Node number 7: 994 observations, complexity param=0.002225189
## predicted class=N expected loss=0.3702213 P(node) =0.2221229
## class counts: 626 368
## probabilities: 0.630 0.370
## left son=14 (85 obs) right son=15 (909 obs)
## Primary splits:
## WordCount < 1944.5 to the right, improve=3.384313, (0 missing)
##
## Node number 12: 73 observations
## predicted class=N expected loss=0.1780822 P(node) =0.01631285
## class counts: 60 13
## probabilities: 0.822 0.178
##
## Node number 13: 117 observations, complexity param=0.001780151
## predicted class=N expected loss=0.3333333 P(node) =0.02614525
## class counts: 78 39
## probabilities: 0.667 0.333
## left son=26 (84 obs) right son=27 (33 obs)
## Primary splits:
## WordCount < 732.5 to the left, improve=0.7597403, (0 missing)
##
## Node number 14: 85 observations, complexity param=0.002225189
## predicted class=N expected loss=0.2352941 P(node) =0.01899441
## class counts: 65 20
## probabilities: 0.765 0.235
## left son=28 (77 obs) right son=29 (8 obs)
## Primary splits:
## WordCount < 3749 to the left, improve=4.679144, (0 missing)
##
## Node number 15: 909 observations, complexity param=0.002225189
## predicted class=N expected loss=0.3828383 P(node) =0.2031285
## class counts: 561 348
## probabilities: 0.617 0.383
## left son=30 (724 obs) right son=31 (185 obs)
## Primary splits:
## WordCount < 875.5 to the right, improve=1.405223, (0 missing)
##
## Node number 26: 84 observations, complexity param=0.001780151
## predicted class=N expected loss=0.297619 P(node) =0.01877095
## class counts: 59 25
## probabilities: 0.702 0.298
## left son=52 (23 obs) right son=53 (61 obs)
## Primary splits:
## WordCount < 719.5 to the right, improve=1.770509, (0 missing)
##
## Node number 27: 33 observations, complexity param=0.001780151
## predicted class=N expected loss=0.4242424 P(node) =0.007374302
## class counts: 19 14
## probabilities: 0.576 0.424
## left son=54 (20 obs) right son=55 (13 obs)
## Primary splits:
## WordCount < 738.5 to the right, improve=1.567366, (0 missing)
##
## Node number 28: 77 observations
## predicted class=N expected loss=0.1818182 P(node) =0.0172067
## class counts: 63 14
## probabilities: 0.818 0.182
##
## Node number 29: 8 observations
## predicted class=Y expected loss=0.25 P(node) =0.001787709
## class counts: 2 6
## probabilities: 0.250 0.750
##
## Node number 30: 724 observations, complexity param=0.002225189
## predicted class=N expected loss=0.3687845 P(node) =0.1617877
## class counts: 457 267
## probabilities: 0.631 0.369
## left son=60 (11 obs) right son=61 (713 obs)
## Primary splits:
## WordCount < 881.5 to the left, improve=3.038205, (0 missing)
##
## Node number 31: 185 observations, complexity param=0.002225189
## predicted class=N expected loss=0.4378378 P(node) =0.04134078
## class counts: 104 81
## probabilities: 0.562 0.438
## left son=62 (175 obs) right son=63 (10 obs)
## Primary splits:
## WordCount < 871.5 to the left, improve=2.773127, (0 missing)
##
## Node number 52: 23 observations
## predicted class=N expected loss=0.1304348 P(node) =0.005139665
## class counts: 20 3
## probabilities: 0.870 0.130
##
## Node number 53: 61 observations, complexity param=0.001780151
## predicted class=N expected loss=0.3606557 P(node) =0.01363128
## class counts: 39 22
## probabilities: 0.639 0.361
## left son=106 (32 obs) right son=107 (29 obs)
## Primary splits:
## WordCount < 701 to the left, improve=0.84882, (0 missing)
##
## Node number 54: 20 observations
## predicted class=N expected loss=0.3 P(node) =0.004469274
## class counts: 14 6
## probabilities: 0.700 0.300
##
## Node number 55: 13 observations
## predicted class=Y expected loss=0.3846154 P(node) =0.002905028
## class counts: 5 8
## probabilities: 0.385 0.615
##
## Node number 60: 11 observations
## predicted class=N expected loss=0 P(node) =0.002458101
## class counts: 11 0
## probabilities: 1.000 0.000
##
## Node number 61: 713 observations, complexity param=0.002225189
## predicted class=N expected loss=0.3744741 P(node) =0.1593296
## class counts: 446 267
## probabilities: 0.626 0.374
## left son=122 (508 obs) right son=123 (205 obs)
## Primary splits:
## WordCount < 1289.5 to the left, improve=1.727744, (0 missing)
##
## Node number 62: 175 observations, complexity param=0.002225189
## predicted class=N expected loss=0.4171429 P(node) =0.03910615
## class counts: 102 73
## probabilities: 0.583 0.417
## left son=124 (124 obs) right son=125 (51 obs)
## Primary splits:
## WordCount < 841.5 to the left, improve=0.7682371, (0 missing)
##
## Node number 63: 10 observations
## predicted class=Y expected loss=0.2 P(node) =0.002234637
## class counts: 2 8
## probabilities: 0.200 0.800
##
## Node number 106: 32 observations
## predicted class=N expected loss=0.28125 P(node) =0.007150838
## class counts: 23 9
## probabilities: 0.719 0.281
##
## Node number 107: 29 observations, complexity param=0.001780151
## predicted class=N expected loss=0.4482759 P(node) =0.006480447
## class counts: 16 13
## probabilities: 0.552 0.448
## left son=214 (18 obs) right son=215 (11 obs)
## Primary splits:
## WordCount < 710 to the right, improve=2.758969, (0 missing)
##
## Node number 122: 508 observations, complexity param=0.001907305
## predicted class=N expected loss=0.3523622 P(node) =0.1135196
## class counts: 329 179
## probabilities: 0.648 0.352
## left son=244 (187 obs) right son=245 (321 obs)
## Primary splits:
## WordCount < 1076.5 to the right, improve=2.812999, (0 missing)
##
## Node number 123: 205 observations, complexity param=0.002225189
## predicted class=N expected loss=0.4292683 P(node) =0.04581006
## class counts: 117 88
## probabilities: 0.571 0.429
## left son=246 (196 obs) right son=247 (9 obs)
## Primary splits:
## WordCount < 1304.5 to the right, improve=2.286649, (0 missing)
##
## Node number 124: 124 observations, complexity param=0.002225189
## predicted class=N expected loss=0.3870968 P(node) =0.0277095
## class counts: 76 48
## probabilities: 0.613 0.387
## left son=248 (40 obs) right son=249 (84 obs)
## Primary splits:
## WordCount < 822.5 to the right, improve=1.483948, (0 missing)
##
## Node number 125: 51 observations, complexity param=0.002225189
## predicted class=N expected loss=0.4901961 P(node) =0.01139665
## class counts: 26 25
## probabilities: 0.510 0.490
## left son=250 (39 obs) right son=251 (12 obs)
## Primary splits:
## WordCount < 849.5 to the right, improve=2.118401, (0 missing)
##
## Node number 214: 18 observations
## predicted class=N expected loss=0.2777778 P(node) =0.004022346
## class counts: 13 5
## probabilities: 0.722 0.278
##
## Node number 215: 11 observations
## predicted class=Y expected loss=0.2727273 P(node) =0.002458101
## class counts: 3 8
## probabilities: 0.273 0.727
##
## Node number 244: 187 observations, complexity param=0.0008900757
## predicted class=N expected loss=0.2834225 P(node) =0.04178771
## class counts: 134 53
## probabilities: 0.717 0.283
## left son=488 (8 obs) right son=489 (179 obs)
## Primary splits:
## WordCount < 1283.5 to the right, improve=1.342694, (0 missing)
##
## Node number 245: 321 observations, complexity param=0.001907305
## predicted class=N expected loss=0.3925234 P(node) =0.07173184
## class counts: 195 126
## probabilities: 0.607 0.393
## left son=490 (229 obs) right son=491 (92 obs)
## Primary splits:
## WordCount < 1004 to the left, improve=1.056392, (0 missing)
##
## Node number 246: 196 observations, complexity param=0.002225189
## predicted class=N expected loss=0.4132653 P(node) =0.04379888
## class counts: 115 81
## probabilities: 0.587 0.413
## left son=492 (111 obs) right son=493 (85 obs)
## Primary splits:
## WordCount < 1443 to the right, improve=0.6230395, (0 missing)
##
## Node number 247: 9 observations
## predicted class=Y expected loss=0.2222222 P(node) =0.002011173
## class counts: 2 7
## probabilities: 0.222 0.778
##
## Node number 248: 40 observations
## predicted class=N expected loss=0.275 P(node) =0.008938547
## class counts: 29 11
## probabilities: 0.725 0.275
##
## Node number 249: 84 observations, complexity param=0.002225189
## predicted class=N expected loss=0.4404762 P(node) =0.01877095
## class counts: 47 37
## probabilities: 0.560 0.440
## left son=498 (35 obs) right son=499 (49 obs)
## Primary splits:
## WordCount < 799.5 to the left, improve=1.143537, (0 missing)
##
## Node number 250: 39 observations
## predicted class=N expected loss=0.4102564 P(node) =0.008715084
## class counts: 23 16
## probabilities: 0.590 0.410
##
## Node number 251: 12 observations
## predicted class=Y expected loss=0.25 P(node) =0.002681564
## class counts: 3 9
## probabilities: 0.250 0.750
##
## Node number 488: 8 observations
## predicted class=N expected loss=0 P(node) =0.001787709
## class counts: 8 0
## probabilities: 1.000 0.000
##
## Node number 489: 179 observations, complexity param=0.0008900757
## predicted class=N expected loss=0.2960894 P(node) =0.04
## class counts: 126 53
## probabilities: 0.704 0.296
## left son=978 (125 obs) right son=979 (54 obs)
## Primary splits:
## WordCount < 1197 to the left, improve=1.916451, (0 missing)
##
## Node number 490: 229 observations, complexity param=0.001907305
## predicted class=N expected loss=0.3668122 P(node) =0.05117318
## class counts: 145 84
## probabilities: 0.633 0.367
## left son=980 (36 obs) right son=981 (193 obs)
## Primary splits:
## WordCount < 983.5 to the right, improve=1.786024, (0 missing)
##
## Node number 491: 92 observations, complexity param=0.001907305
## predicted class=N expected loss=0.4565217 P(node) =0.02055866
## class counts: 50 42
## probabilities: 0.543 0.457
## left son=982 (61 obs) right son=983 (31 obs)
## Primary splits:
## WordCount < 1027.5 to the right, improve=0.7891385, (0 missing)
##
## Node number 492: 111 observations, complexity param=0.001335113
## predicted class=N expected loss=0.3783784 P(node) =0.02480447
## class counts: 69 42
## probabilities: 0.622 0.378
## left son=984 (34 obs) right son=985 (77 obs)
## Primary splits:
## WordCount < 1548.5 to the left, improve=1.266636, (0 missing)
##
## Node number 493: 85 observations, complexity param=0.002225189
## predicted class=N expected loss=0.4588235 P(node) =0.01899441
## class counts: 46 39
## probabilities: 0.541 0.459
## left son=986 (69 obs) right son=987 (16 obs)
## Primary splits:
## WordCount < 1417.5 to the left, improve=1.088576, (0 missing)
##
## Node number 498: 35 observations
## predicted class=N expected loss=0.3428571 P(node) =0.007821229
## class counts: 23 12
## probabilities: 0.657 0.343
##
## Node number 499: 49 observations, complexity param=0.002225189
## predicted class=Y expected loss=0.4897959 P(node) =0.01094972
## class counts: 24 25
## probabilities: 0.490 0.510
## left son=998 (40 obs) right son=999 (9 obs)
## Primary splits:
## WordCount < 804.5 to the right, improve=1.578685, (0 missing)
##
## Node number 978: 125 observations, complexity param=0.0002670227
## predicted class=N expected loss=0.248 P(node) =0.02793296
## class counts: 94 31
## probabilities: 0.752 0.248
## left son=1956 (9 obs) right son=1957 (116 obs)
## Primary splits:
## WordCount < 1187 to the right, improve=1.192966, (0 missing)
##
## Node number 979: 54 observations, complexity param=0.0008900757
## predicted class=N expected loss=0.4074074 P(node) =0.01206704
## class counts: 32 22
## probabilities: 0.593 0.407
## left son=1958 (42 obs) right son=1959 (12 obs)
## Primary splits:
## WordCount < 1211 to the right, improve=0.9550265, (0 missing)
##
## Node number 980: 36 observations
## predicted class=N expected loss=0.2222222 P(node) =0.008044693
## class counts: 28 8
## probabilities: 0.778 0.222
##
## Node number 981: 193 observations, complexity param=0.001907305
## predicted class=N expected loss=0.3937824 P(node) =0.04312849
## class counts: 117 76
## probabilities: 0.606 0.394
## left son=1962 (183 obs) right son=1963 (10 obs)
## Primary splits:
## WordCount < 976 to the left, improve=0.8969903, (0 missing)
##
## Node number 982: 61 observations, complexity param=0.001907305
## predicted class=N expected loss=0.4098361 P(node) =0.01363128
## class counts: 36 25
## probabilities: 0.590 0.410
## left son=1964 (10 obs) right son=1965 (51 obs)
## Primary splits:
## WordCount < 1034 to the left, improve=2.296432, (0 missing)
##
## Node number 983: 31 observations, complexity param=0.001907305
## predicted class=Y expected loss=0.4516129 P(node) =0.006927374
## class counts: 14 17
## probabilities: 0.452 0.548
## left son=1966 (24 obs) right son=1967 (7 obs)
## Primary splits:
## WordCount < 1019.5 to the left, improve=0.4976959, (0 missing)
##
## Node number 984: 34 observations
## predicted class=N expected loss=0.2647059 P(node) =0.007597765
## class counts: 25 9
## probabilities: 0.735 0.265
##
## Node number 985: 77 observations, complexity param=0.001335113
## predicted class=N expected loss=0.4285714 P(node) =0.0172067
## class counts: 44 33
## probabilities: 0.571 0.429
## left son=1970 (36 obs) right son=1971 (41 obs)
## Primary splits:
## WordCount < 1709.5 to the right, improve=0.2129307, (0 missing)
##
## Node number 986: 69 observations, complexity param=0.00200267
## predicted class=N expected loss=0.4202899 P(node) =0.01541899
## class counts: 40 29
## probabilities: 0.580 0.420
## left son=1972 (14 obs) right son=1973 (55 obs)
## Primary splits:
## WordCount < 1327.5 to the left, improve=0.6361754, (0 missing)
##
## Node number 987: 16 observations
## predicted class=Y expected loss=0.375 P(node) =0.003575419
## class counts: 6 10
## probabilities: 0.375 0.625
##
## Node number 998: 40 observations, complexity param=0.002225189
## predicted class=N expected loss=0.45 P(node) =0.008938547
## class counts: 22 18
## probabilities: 0.550 0.450
## left son=1996 (11 obs) right son=1997 (29 obs)
## Primary splits:
## WordCount < 810.5 to the left, improve=2.182445, (0 missing)
##
## Node number 999: 9 observations
## predicted class=Y expected loss=0.2222222 P(node) =0.002011173
## class counts: 2 7
## probabilities: 0.222 0.778
##
## Node number 1956: 9 observations
## predicted class=N expected loss=0 P(node) =0.002011173
## class counts: 9 0
## probabilities: 1.000 0.000
##
## Node number 1957: 116 observations, complexity param=0.0002670227
## predicted class=N expected loss=0.2672414 P(node) =0.02592179
## class counts: 85 31
## probabilities: 0.733 0.267
## left son=3914 (96 obs) right son=3915 (20 obs)
## Primary splits:
## WordCount < 1164 to the left, improve=0.8518678, (0 missing)
##
## Node number 1958: 42 observations, complexity param=0.0008900757
## predicted class=N expected loss=0.3571429 P(node) =0.009385475
## class counts: 27 15
## probabilities: 0.643 0.357
## left son=3916 (10 obs) right son=3917 (32 obs)
## Primary splits:
## WordCount < 1225.5 to the left, improve=0.6482143, (0 missing)
##
## Node number 1959: 12 observations
## predicted class=Y expected loss=0.4166667 P(node) =0.002681564
## class counts: 5 7
## probabilities: 0.417 0.583
##
## Node number 1962: 183 observations, complexity param=0.001907305
## predicted class=N expected loss=0.3825137 P(node) =0.04089385
## class counts: 113 70
## probabilities: 0.617 0.383
## left son=3924 (18 obs) right son=3925 (165 obs)
## Primary splits:
## WordCount < 965 to the right, improve=1.025865, (0 missing)
##
## Node number 1963: 10 observations
## predicted class=Y expected loss=0.4 P(node) =0.002234637
## class counts: 4 6
## probabilities: 0.400 0.600
##
## Node number 1964: 10 observations
## predicted class=N expected loss=0.1 P(node) =0.002234637
## class counts: 9 1
## probabilities: 0.900 0.100
##
## Node number 1965: 51 observations, complexity param=0.001907305
## predicted class=N expected loss=0.4705882 P(node) =0.01139665
## class counts: 27 24
## probabilities: 0.529 0.471
## left son=3930 (23 obs) right son=3931 (28 obs)
## Primary splits:
## WordCount < 1056 to the right, improve=0.5266715, (0 missing)
##
## Node number 1966: 24 observations, complexity param=0.001907305
## predicted class=N expected loss=0.5 P(node) =0.005363128
## class counts: 12 12
## probabilities: 0.500 0.500
## left son=3932 (9 obs) right son=3933 (15 obs)
## Primary splits:
## WordCount < 1011.5 to the right, improve=0.8, (0 missing)
##
## Node number 1967: 7 observations
## predicted class=Y expected loss=0.2857143 P(node) =0.001564246
## class counts: 2 5
## probabilities: 0.286 0.714
##
## Node number 1970: 36 observations
## predicted class=N expected loss=0.3888889 P(node) =0.008044693
## class counts: 22 14
## probabilities: 0.611 0.389
##
## Node number 1971: 41 observations, complexity param=0.001335113
## predicted class=N expected loss=0.4634146 P(node) =0.009162011
## class counts: 22 19
## probabilities: 0.537 0.463
## left son=3942 (26 obs) right son=3943 (15 obs)
## Primary splits:
## WordCount < 1644 to the left, improve=0.8825516, (0 missing)
##
## Node number 1972: 14 observations
## predicted class=N expected loss=0.2857143 P(node) =0.003128492
## class counts: 10 4
## probabilities: 0.714 0.286
##
## Node number 1973: 55 observations, complexity param=0.00200267
## predicted class=N expected loss=0.4545455 P(node) =0.0122905
## class counts: 30 25
## probabilities: 0.545 0.455
## left son=3946 (40 obs) right son=3947 (15 obs)
## Primary splits:
## WordCount < 1350.5 to the right, improve=0.8727273, (0 missing)
##
## Node number 1996: 11 observations
## predicted class=N expected loss=0.1818182 P(node) =0.002458101
## class counts: 9 2
## probabilities: 0.818 0.182
##
## Node number 1997: 29 observations
## predicted class=Y expected loss=0.4482759 P(node) =0.006480447
## class counts: 13 16
## probabilities: 0.448 0.552
##
## Node number 3914: 96 observations, complexity param=0.0002670227
## predicted class=N expected loss=0.2395833 P(node) =0.02145251
## class counts: 73 23
## probabilities: 0.760 0.240
## left son=7828 (26 obs) right son=7829 (70 obs)
## Primary splits:
## WordCount < 1145.5 to the right, improve=0.5242216, (0 missing)
##
## Node number 3915: 20 observations
## predicted class=N expected loss=0.4 P(node) =0.004469274
## class counts: 12 8
## probabilities: 0.600 0.400
##
## Node number 3916: 10 observations
## predicted class=N expected loss=0.2 P(node) =0.002234637
## class counts: 8 2
## probabilities: 0.800 0.200
##
## Node number 3917: 32 observations, complexity param=0.0008900757
## predicted class=N expected loss=0.40625 P(node) =0.007150838
## class counts: 19 13
## probabilities: 0.594 0.406
## left son=7834 (22 obs) right son=7835 (10 obs)
## Primary splits:
## WordCount < 1245 to the right, improve=1.092045, (0 missing)
##
## Node number 3924: 18 observations
## predicted class=N expected loss=0.2222222 P(node) =0.004022346
## class counts: 14 4
## probabilities: 0.778 0.222
##
## Node number 3925: 165 observations, complexity param=0.001907305
## predicted class=N expected loss=0.4 P(node) =0.03687151
## class counts: 99 66
## probabilities: 0.600 0.400
## left son=7850 (150 obs) right son=7851 (15 obs)
## Primary splits:
## WordCount < 955 to the left, improve=2.346667, (0 missing)
##
## Node number 3930: 23 observations
## predicted class=N expected loss=0.3913043 P(node) =0.005139665
## class counts: 14 9
## probabilities: 0.609 0.391
##
## Node number 3931: 28 observations, complexity param=0.001907305
## predicted class=Y expected loss=0.4642857 P(node) =0.006256983
## class counts: 13 15
## probabilities: 0.464 0.536
## left son=7862 (21 obs) right son=7863 (7 obs)
## Primary splits:
## WordCount < 1050 to the left, improve=1.928571, (0 missing)
##
## Node number 3932: 9 observations
## predicted class=N expected loss=0.3333333 P(node) =0.002011173
## class counts: 6 3
## probabilities: 0.667 0.333
##
## Node number 3933: 15 observations
## predicted class=Y expected loss=0.4 P(node) =0.003351955
## class counts: 6 9
## probabilities: 0.400 0.600
##
## Node number 3942: 26 observations
## predicted class=N expected loss=0.3846154 P(node) =0.005810056
## class counts: 16 10
## probabilities: 0.615 0.385
##
## Node number 3943: 15 observations
## predicted class=Y expected loss=0.4 P(node) =0.003351955
## class counts: 6 9
## probabilities: 0.400 0.600
##
## Node number 3946: 40 observations, complexity param=0.001335113
## predicted class=N expected loss=0.4 P(node) =0.008938547
## class counts: 24 16
## probabilities: 0.600 0.400
## left son=7892 (7 obs) right son=7893 (33 obs)
## Primary splits:
## WordCount < 1366 to the left, improve=1.122078, (0 missing)
##
## Node number 3947: 15 observations
## predicted class=Y expected loss=0.4 P(node) =0.003351955
## class counts: 6 9
## probabilities: 0.400 0.600
##
## Node number 7828: 26 observations
## predicted class=N expected loss=0.1538462 P(node) =0.005810056
## class counts: 22 4
## probabilities: 0.846 0.154
##
## Node number 7829: 70 observations, complexity param=0.0002670227
## predicted class=N expected loss=0.2714286 P(node) =0.01564246
## class counts: 51 19
## probabilities: 0.729 0.271
## left son=15658 (30 obs) right son=15659 (40 obs)
## Primary splits:
## WordCount < 1101 to the left, improve=0.5357143, (0 missing)
##
## Node number 7834: 22 observations
## predicted class=N expected loss=0.3181818 P(node) =0.004916201
## class counts: 15 7
## probabilities: 0.682 0.318
##
## Node number 7835: 10 observations
## predicted class=Y expected loss=0.4 P(node) =0.002234637
## class counts: 4 6
## probabilities: 0.400 0.600
##
## Node number 7850: 150 observations, complexity param=0.001907305
## predicted class=N expected loss=0.3733333 P(node) =0.03351955
## class counts: 94 56
## probabilities: 0.627 0.373
## left son=15700 (32 obs) right son=15701 (118 obs)
## Primary splits:
## WordCount < 936.5 to the right, improve=1.237514, (0 missing)
##
## Node number 7851: 15 observations
## predicted class=Y expected loss=0.3333333 P(node) =0.003351955
## class counts: 5 10
## probabilities: 0.333 0.667
##
## Node number 7862: 21 observations
## predicted class=N expected loss=0.4285714 P(node) =0.004692737
## class counts: 12 9
## probabilities: 0.571 0.429
##
## Node number 7863: 7 observations
## predicted class=Y expected loss=0.1428571 P(node) =0.001564246
## class counts: 1 6
## probabilities: 0.143 0.857
##
## Node number 7892: 7 observations
## predicted class=N expected loss=0.1428571 P(node) =0.001564246
## class counts: 6 1
## probabilities: 0.857 0.143
##
## Node number 7893: 33 observations, complexity param=0.001335113
## predicted class=N expected loss=0.4545455 P(node) =0.007374302
## class counts: 18 15
## probabilities: 0.545 0.455
## left son=15786 (15 obs) right son=15787 (18 obs)
## Primary splits:
## WordCount < 1391 to the right, improve=0.8080808, (0 missing)
##
## Node number 15658: 30 observations
## predicted class=N expected loss=0.2 P(node) =0.006703911
## class counts: 24 6
## probabilities: 0.800 0.200
##
## Node number 15659: 40 observations, complexity param=0.0002670227
## predicted class=N expected loss=0.325 P(node) =0.008938547
## class counts: 27 13
## probabilities: 0.675 0.325
## left son=31318 (29 obs) right son=31319 (11 obs)
## Primary splits:
## WordCount < 1110 to the right, improve=1.474765, (0 missing)
##
## Node number 15700: 32 observations
## predicted class=N expected loss=0.25 P(node) =0.007150838
## class counts: 24 8
## probabilities: 0.750 0.250
##
## Node number 15701: 118 observations, complexity param=0.001907305
## predicted class=N expected loss=0.4067797 P(node) =0.02636872
## class counts: 70 48
## probabilities: 0.593 0.407
## left son=31402 (103 obs) right son=31403 (15 obs)
## Primary splits:
## WordCount < 930.5 to the left, improve=1.283133, (0 missing)
##
## Node number 15786: 15 observations
## predicted class=N expected loss=0.3333333 P(node) =0.003351955
## class counts: 10 5
## probabilities: 0.667 0.333
##
## Node number 15787: 18 observations
## predicted class=Y expected loss=0.4444444 P(node) =0.004022346
## class counts: 8 10
## probabilities: 0.444 0.556
##
## Node number 31318: 29 observations
## predicted class=N expected loss=0.2413793 P(node) =0.006480447
## class counts: 22 7
## probabilities: 0.759 0.241
##
## Node number 31319: 11 observations
## predicted class=Y expected loss=0.4545455 P(node) =0.002458101
## class counts: 5 6
## probabilities: 0.455 0.545
##
## Node number 31402: 103 observations, complexity param=0.001335113
## predicted class=N expected loss=0.3786408 P(node) =0.02301676
## class counts: 64 39
## probabilities: 0.621 0.379
## left son=62804 (75 obs) right son=62805 (28 obs)
## Primary splits:
## WordCount < 890.5 to the right, improve=1.132686, (0 missing)
##
## Node number 31403: 15 observations
## predicted class=Y expected loss=0.4 P(node) =0.003351955
## class counts: 6 9
## probabilities: 0.400 0.600
##
## Node number 62804: 75 observations, complexity param=0.001335113
## predicted class=N expected loss=0.3333333 P(node) =0.01675978
## class counts: 50 25
## probabilities: 0.667 0.333
## left son=125608 (11 obs) right son=125609 (64 obs)
## Primary splits:
## WordCount < 923.5 to the right, improve=0.5918561, (0 missing)
##
## Node number 62805: 28 observations, complexity param=0.001335113
## predicted class=N expected loss=0.5 P(node) =0.006256983
## class counts: 14 14
## probabilities: 0.500 0.500
## left son=125610 (7 obs) right son=125611 (21 obs)
## Primary splits:
## WordCount < 888.5 to the right, improve=0.0952381, (0 missing)
##
## Node number 125608: 11 observations
## predicted class=N expected loss=0.1818182 P(node) =0.002458101
## class counts: 9 2
## probabilities: 0.818 0.182
##
## Node number 125609: 64 observations, complexity param=0.001335113
## predicted class=N expected loss=0.359375 P(node) =0.01430168
## class counts: 41 23
## probabilities: 0.641 0.359
## left son=251218 (21 obs) right son=251219 (43 obs)
## Primary splits:
## WordCount < 903.5 to the left, improve=0.9194698, (0 missing)
##
## Node number 125610: 7 observations
## predicted class=N expected loss=0.4285714 P(node) =0.001564246
## class counts: 4 3
## probabilities: 0.571 0.429
##
## Node number 125611: 21 observations
## predicted class=Y expected loss=0.4761905 P(node) =0.004692737
## class counts: 10 11
## probabilities: 0.476 0.524
##
## Node number 251218: 21 observations
## predicted class=N expected loss=0.2380952 P(node) =0.004692737
## class counts: 16 5
## probabilities: 0.762 0.238
##
## Node number 251219: 43 observations, complexity param=0.001335113
## predicted class=N expected loss=0.4186047 P(node) =0.009608939
## class counts: 25 18
## probabilities: 0.581 0.419
## left son=502438 (35 obs) right son=502439 (8 obs)
## Primary splits:
## WordCount < 906.5 to the right, improve=2.158804, (0 missing)
##
## Node number 502438: 35 observations
## predicted class=N expected loss=0.3428571 P(node) =0.007821229
## class counts: 23 12
## probabilities: 0.657 0.343
##
## Node number 502439: 8 observations
## predicted class=Y expected loss=0.25 P(node) =0.001787709
## class counts: 2 6
## probabilities: 0.250 0.750
##
## n= 4475
##
## node), split, n, loss, yval, (yprob)
## * denotes terminal node
##
## 1) root 4475 749 N (0.83262570 0.16737430)
## 2) WordCount< 683.5 3291 329 N (0.90003039 0.09996961) *
## 3) WordCount>=683.5 1184 420 N (0.64527027 0.35472973)
## 6) WordCount< 782.5 190 52 N (0.72631579 0.27368421)
## 12) WordCount>=749.5 73 13 N (0.82191781 0.17808219) *
## 13) WordCount< 749.5 117 39 N (0.66666667 0.33333333)
## 26) WordCount< 732.5 84 25 N (0.70238095 0.29761905)
## 52) WordCount>=719.5 23 3 N (0.86956522 0.13043478) *
## 53) WordCount< 719.5 61 22 N (0.63934426 0.36065574)
## 106) WordCount< 701 32 9 N (0.71875000 0.28125000) *
## 107) WordCount>=701 29 13 N (0.55172414 0.44827586)
## 214) WordCount>=710 18 5 N (0.72222222 0.27777778) *
## 215) WordCount< 710 11 3 Y (0.27272727 0.72727273) *
## 27) WordCount>=732.5 33 14 N (0.57575758 0.42424242)
## 54) WordCount>=738.5 20 6 N (0.70000000 0.30000000) *
## 55) WordCount< 738.5 13 5 Y (0.38461538 0.61538462) *
## 7) WordCount>=782.5 994 368 N (0.62977867 0.37022133)
## 14) WordCount>=1944.5 85 20 N (0.76470588 0.23529412)
## 28) WordCount< 3749 77 14 N (0.81818182 0.18181818) *
## 29) WordCount>=3749 8 2 Y (0.25000000 0.75000000) *
## 15) WordCount< 1944.5 909 348 N (0.61716172 0.38283828)
## 30) WordCount>=875.5 724 267 N (0.63121547 0.36878453)
## 60) WordCount< 881.5 11 0 N (1.00000000 0.00000000) *
## 61) WordCount>=881.5 713 267 N (0.62552595 0.37447405)
## 122) WordCount< 1289.5 508 179 N (0.64763780 0.35236220)
## 244) WordCount>=1076.5 187 53 N (0.71657754 0.28342246)
## 488) WordCount>=1283.5 8 0 N (1.00000000 0.00000000) *
## 489) WordCount< 1283.5 179 53 N (0.70391061 0.29608939)
## 978) WordCount< 1197 125 31 N (0.75200000 0.24800000)
## 1956) WordCount>=1187 9 0 N (1.00000000 0.00000000) *
## 1957) WordCount< 1187 116 31 N (0.73275862 0.26724138)
## 3914) WordCount< 1164 96 23 N (0.76041667 0.23958333)
## 7828) WordCount>=1145.5 26 4 N (0.84615385 0.15384615) *
## 7829) WordCount< 1145.5 70 19 N (0.72857143 0.27142857)
## 15658) WordCount< 1101 30 6 N (0.80000000 0.20000000) *
## 15659) WordCount>=1101 40 13 N (0.67500000 0.32500000)
## 31318) WordCount>=1110 29 7 N (0.75862069 0.24137931) *
## 31319) WordCount< 1110 11 5 Y (0.45454545 0.54545455) *
## 3915) WordCount>=1164 20 8 N (0.60000000 0.40000000) *
## 979) WordCount>=1197 54 22 N (0.59259259 0.40740741)
## 1958) WordCount>=1211 42 15 N (0.64285714 0.35714286)
## 3916) WordCount< 1225.5 10 2 N (0.80000000 0.20000000) *
## 3917) WordCount>=1225.5 32 13 N (0.59375000 0.40625000)
## 7834) WordCount>=1245 22 7 N (0.68181818 0.31818182) *
## 7835) WordCount< 1245 10 4 Y (0.40000000 0.60000000) *
## 1959) WordCount< 1211 12 5 Y (0.41666667 0.58333333) *
## 245) WordCount< 1076.5 321 126 N (0.60747664 0.39252336)
## 490) WordCount< 1004 229 84 N (0.63318777 0.36681223)
## 980) WordCount>=983.5 36 8 N (0.77777778 0.22222222) *
## 981) WordCount< 983.5 193 76 N (0.60621762 0.39378238)
## 1962) WordCount< 976 183 70 N (0.61748634 0.38251366)
## 3924) WordCount>=965 18 4 N (0.77777778 0.22222222) *
## 3925) WordCount< 965 165 66 N (0.60000000 0.40000000)
## 7850) WordCount< 955 150 56 N (0.62666667 0.37333333)
## 15700) WordCount>=936.5 32 8 N (0.75000000 0.25000000) *
## 15701) WordCount< 936.5 118 48 N (0.59322034 0.40677966)
## 31402) WordCount< 930.5 103 39 N (0.62135922 0.37864078)
## 62804) WordCount>=890.5 75 25 N (0.66666667 0.33333333)
## 125608) WordCount>=923.5 11 2 N (0.81818182 0.18181818) *
## 125609) WordCount< 923.5 64 23 N (0.64062500 0.35937500)
## 251218) WordCount< 903.5 21 5 N (0.76190476 0.23809524) *
## 251219) WordCount>=903.5 43 18 N (0.58139535 0.41860465)
## 502438) WordCount>=906.5 35 12 N (0.65714286 0.34285714) *
## 502439) WordCount< 906.5 8 2 Y (0.25000000 0.75000000) *
## 62805) WordCount< 890.5 28 14 N (0.50000000 0.50000000)
## 125610) WordCount>=888.5 7 3 N (0.57142857 0.42857143) *
## 125611) WordCount< 888.5 21 10 Y (0.47619048 0.52380952) *
## 31403) WordCount>=930.5 15 6 Y (0.40000000 0.60000000) *
## 7851) WordCount>=955 15 5 Y (0.33333333 0.66666667) *
## 1963) WordCount>=976 10 4 Y (0.40000000 0.60000000) *
## 491) WordCount>=1004 92 42 N (0.54347826 0.45652174)
## 982) WordCount>=1027.5 61 25 N (0.59016393 0.40983607)
## 1964) WordCount< 1034 10 1 N (0.90000000 0.10000000) *
## 1965) WordCount>=1034 51 24 N (0.52941176 0.47058824)
## 3930) WordCount>=1056 23 9 N (0.60869565 0.39130435) *
## 3931) WordCount< 1056 28 13 Y (0.46428571 0.53571429)
## 7862) WordCount< 1050 21 9 N (0.57142857 0.42857143) *
## 7863) WordCount>=1050 7 1 Y (0.14285714 0.85714286) *
## 983) WordCount< 1027.5 31 14 Y (0.45161290 0.54838710)
## 1966) WordCount< 1019.5 24 12 N (0.50000000 0.50000000)
## 3932) WordCount>=1011.5 9 3 N (0.66666667 0.33333333) *
## 3933) WordCount< 1011.5 15 6 Y (0.40000000 0.60000000) *
## 1967) WordCount>=1019.5 7 2 Y (0.28571429 0.71428571) *
## 123) WordCount>=1289.5 205 88 N (0.57073171 0.42926829)
## 246) WordCount>=1304.5 196 81 N (0.58673469 0.41326531)
## 492) WordCount>=1443 111 42 N (0.62162162 0.37837838)
## 984) WordCount< 1548.5 34 9 N (0.73529412 0.26470588) *
## 985) WordCount>=1548.5 77 33 N (0.57142857 0.42857143)
## 1970) WordCount>=1709.5 36 14 N (0.61111111 0.38888889) *
## 1971) WordCount< 1709.5 41 19 N (0.53658537 0.46341463)
## 3942) WordCount< 1644 26 10 N (0.61538462 0.38461538) *
## 3943) WordCount>=1644 15 6 Y (0.40000000 0.60000000) *
## 493) WordCount< 1443 85 39 N (0.54117647 0.45882353)
## 986) WordCount< 1417.5 69 29 N (0.57971014 0.42028986)
## 1972) WordCount< 1327.5 14 4 N (0.71428571 0.28571429) *
## 1973) WordCount>=1327.5 55 25 N (0.54545455 0.45454545)
## 3946) WordCount>=1350.5 40 16 N (0.60000000 0.40000000)
## 7892) WordCount< 1366 7 1 N (0.85714286 0.14285714) *
## 7893) WordCount>=1366 33 15 N (0.54545455 0.45454545)
## 15786) WordCount>=1391 15 5 N (0.66666667 0.33333333) *
## 15787) WordCount< 1391 18 8 Y (0.44444444 0.55555556) *
## 3947) WordCount< 1350.5 15 6 Y (0.40000000 0.60000000) *
## 987) WordCount>=1417.5 16 6 Y (0.37500000 0.62500000) *
## 247) WordCount< 1304.5 9 2 Y (0.22222222 0.77777778) *
## 31) WordCount< 875.5 185 81 N (0.56216216 0.43783784)
## 62) WordCount< 871.5 175 73 N (0.58285714 0.41714286)
## 124) WordCount< 841.5 124 48 N (0.61290323 0.38709677)
## 248) WordCount>=822.5 40 11 N (0.72500000 0.27500000) *
## 249) WordCount< 822.5 84 37 N (0.55952381 0.44047619)
## 498) WordCount< 799.5 35 12 N (0.65714286 0.34285714) *
## 499) WordCount>=799.5 49 24 Y (0.48979592 0.51020408)
## 998) WordCount>=804.5 40 18 N (0.55000000 0.45000000)
## 1996) WordCount< 810.5 11 2 N (0.81818182 0.18181818) *
## 1997) WordCount>=810.5 29 13 Y (0.44827586 0.55172414) *
## 999) WordCount< 804.5 9 2 Y (0.22222222 0.77777778) *
## 125) WordCount>=841.5 51 25 N (0.50980392 0.49019608)
## 250) WordCount>=849.5 39 16 N (0.58974359 0.41025641) *
## 251) WordCount< 849.5 12 3 Y (0.25000000 0.75000000) *
## 63) WordCount>=871.5 10 2 Y (0.20000000 0.80000000) *
## [1] " calling mypredict_mdl for fit:"
## Reference
## Prediction N Y
## N 0 0
## Y 3726 749
## Popular.fctr Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.N
## 1 N 0
## 2 Y 0
## Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.Y
## 1 3726
## 2 749
## Reference
## Prediction N Y
## N 2999 330
## Y 727 419
## Popular.fctr Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.N
## 1 N 2999
## 2 Y 330
## Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.Y
## 1 727
## 2 419
## Reference
## Prediction N Y
## N 3196 371
## Y 530 378
## Popular.fctr Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.N
## 1 N 3196
## 2 Y 371
## Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.Y
## 1 530
## 2 378
## Reference
## Prediction N Y
## N 3424 447
## Y 302 302
## Popular.fctr Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.N
## 1 N 3424
## 2 Y 447
## Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.Y
## 1 302
## 2 302
## Reference
## Prediction N Y
## N 3579 533
## Y 147 216
## Popular.fctr Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.N
## 1 N 3579
## 2 Y 533
## Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.Y
## 1 147
## 2 216
## Reference
## Prediction N Y
## N 3618 561
## Y 108 188
## Popular.fctr Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.N
## 1 N 3618
## 2 Y 561
## Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.Y
## 1 108
## 2 188
## Reference
## Prediction N Y
## N 3667 623
## Y 59 126
## Popular.fctr Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.N
## 1 N 3667
## 2 Y 623
## Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.Y
## 1 59
## 2 126
## Reference
## Prediction N Y
## N 3707 687
## Y 19 62
## Popular.fctr Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.N
## 1 N 3707
## 2 Y 687
## Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.Y
## 1 19
## 2 62
## Reference
## Prediction N Y
## N 3725 743
## Y 1 6
## Popular.fctr Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.N
## 1 N 3725
## 2 Y 743
## Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.Y
## 1 1
## 2 6
## Reference
## Prediction N Y
## N 3726 749
## Y 0 0
## Popular.fctr Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.N
## 1 N 3726
## 2 Y 749
## Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.Y
## 1 0
## 2 0
## Reference
## Prediction N Y
## N 3726 749
## Y 0 0
## Popular.fctr Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.N
## 1 N 3726
## 2 Y 749
## Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.Y
## 1 0
## 2 0
## threshold f.score
## 1 0.0 0.28675345
## 2 0.1 0.44221636
## 3 0.2 0.45624623
## 4 0.3 0.44641537
## 5 0.4 0.38848921
## 6 0.5 0.35980861
## 7 0.6 0.26980728
## 8 0.7 0.14939759
## 9 0.8 0.01587302
## 10 0.9 0.00000000
## 11 1.0 0.00000000
## [1] "Classifier Probability Threshold: 0.2000 to maximize f.score.fit"
## Popular.fctr Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.N
## 1 N 3196
## 2 Y 371
## Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.Y
## 1 530
## 2 378
## Reference
## Prediction N Y
## N 3196 371
## Y 530 378
## Popular.fctr Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.N
## 1 N 3196
## 2 Y 371
## Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.Y
## 1 530
## 2 378
## [1] "in mypredict_mdl: calling confusionMatrix..."
## [1] " unique(df[, rsp_var_out]: "
## [1] N Y
## Levels: N Y
## [1] " unique(df[, rsp_var ]: "
## [1] Y N
## Levels: N Y
## Prediction
## Reference N Y
## N 3196 530
## Y 371 378
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 7.986592e-01 3.340962e-01 7.866053e-01 8.103220e-01 8.326257e-01
## AccuracyPValue McnemarPValue
## 1.000000e+00 1.411516e-07
## [1] " calling mypredict_mdl for OOB:"
## Reference
## Prediction N Y
## N 0 0
## Y 1713 344
## Popular.fctr Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.N
## 1 N 0
## 2 Y 0
## Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.Y
## 1 1713
## 2 344
## Reference
## Prediction N Y
## N 1350 164
## Y 363 180
## Popular.fctr Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.N
## 1 N 1350
## 2 Y 164
## Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.Y
## 1 363
## 2 180
## Reference
## Prediction N Y
## N 1442 195
## Y 271 149
## Popular.fctr Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.N
## 1 N 1442
## 2 Y 195
## Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.Y
## 1 271
## 2 149
## Reference
## Prediction N Y
## N 1532 245
## Y 181 99
## Popular.fctr Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.N
## 1 N 1532
## 2 Y 245
## Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.Y
## 1 181
## 2 99
## Reference
## Prediction N Y
## N 1610 292
## Y 103 52
## Popular.fctr Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.N
## 1 N 1610
## 2 Y 292
## Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.Y
## 1 103
## 2 52
## Reference
## Prediction N Y
## N 1629 300
## Y 84 44
## Popular.fctr Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.N
## 1 N 1629
## 2 Y 300
## Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.Y
## 1 84
## 2 44
## Reference
## Prediction N Y
## N 1652 310
## Y 61 34
## Popular.fctr Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.N
## 1 N 1652
## 2 Y 310
## Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.Y
## 1 61
## 2 34
## Reference
## Prediction N Y
## N 1681 326
## Y 32 18
## Popular.fctr Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.N
## 1 N 1681
## 2 Y 326
## Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.Y
## 1 32
## 2 18
## Reference
## Prediction N Y
## N 1710 343
## Y 3 1
## Popular.fctr Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.N
## 1 N 1710
## 2 Y 343
## Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.Y
## 1 3
## 2 1
## Reference
## Prediction N Y
## N 1713 344
## Y 0 0
## Popular.fctr Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.N
## 1 N 1713
## 2 Y 344
## Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.Y
## 1 0
## 2 0
## Reference
## Prediction N Y
## N 1713 344
## Y 0 0
## Popular.fctr Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.N
## 1 N 1713
## 2 Y 344
## Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.Y
## 1 0
## 2 0
## threshold f.score
## 1 0.0 0.286547272
## 2 0.1 0.405862458
## 3 0.2 0.390052356
## 4 0.3 0.317307692
## 5 0.4 0.208416834
## 6 0.5 0.186440678
## 7 0.6 0.154897494
## 8 0.7 0.091370558
## 9 0.8 0.005747126
## 10 0.9 0.000000000
## 11 1.0 0.000000000
## [1] "Classifier Probability Threshold: 0.1000 to maximize f.score.OOB"
## Popular.fctr Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.N
## 1 N 1350
## 2 Y 164
## Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.Y
## 1 363
## 2 180
## Reference
## Prediction N Y
## N 1350 164
## Y 363 180
## Popular.fctr Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.N
## 1 N 1350
## 2 Y 164
## Popular.fctr.predict.Max.cor.Y.cv.0.cp.0.rpart.Y
## 1 363
## 2 180
## [1] "in mypredict_mdl: calling confusionMatrix..."
## [1] " unique(df[, rsp_var_out]: "
## [1] Y N
## Levels: N Y
## [1] " unique(df[, rsp_var ]: "
## [1] N Y
## Levels: N Y
## Prediction
## Reference N Y
## N 1350 363
## Y 164 180
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 7.438017e-01 2.528893e-01 7.243549e-01 7.625511e-01 8.327662e-01
## AccuracyPValue McnemarPValue
## 1.000000e+00 6.408399e-18
## model_id model_method feats max.nTuningRuns
## 1 Max.cor.Y.cv.0.cp.0.rpart rpart WordCount 0
## min.elapsedtime.everything min.elapsedtime.final max.auc.fit
## 1 0.599 0.054 0.7072558
## opt.prob.threshold.fit max.f.score.fit max.Accuracy.fit
## 1 0.2 0.4562462 0.7986592
## max.AccuracyLower.fit max.AccuracyUpper.fit max.Kappa.fit max.auc.OOB
## 1 0.7866053 0.810322 0.3340962 0.6552144
## opt.prob.threshold.OOB max.f.score.OOB max.Accuracy.OOB
## 1 0.1 0.4058625 0.7438017
## max.AccuracyLower.OOB max.AccuracyUpper.OOB max.Kappa.OOB
## 1 0.7243549 0.7625511 0.2528893if (glb_is_regression || glb_is_binomial) # For multinomials this model will be run next by default
ret_lst <- myfit_mdl(model_id="Max.cor.Y",
model_method="rpart",
model_type=glb_model_type,
indep_vars_vctr=max_cor_y_x_var,
rsp_var=glb_rsp_var, rsp_var_out=glb_rsp_var_out,
fit_df=glb_fitent_df, OOB_df=glb_OOBent_df,
n_cv_folds=glb_n_cv_folds, tune_models_df=NULL)## [1] "fitting model: Max.cor.Y.rpart"
## [1] " indep_vars: WordCount"
## + Fold1: cp=0.001907
## - Fold1: cp=0.001907
## + Fold2: cp=0.001907
## - Fold2: cp=0.001907
## + Fold3: cp=0.001907
## - Fold3: cp=0.001907
## Aggregating results
## Selecting tuning parameters
## Fitting cp = 0.00223 on full training set## Warning in myfit_mdl(model_id = "Max.cor.Y", model_method = "rpart",
## model_type = glb_model_type, : model's bestTune found at an extreme of
## tuneGrid for parameter: cp
## Call:
## rpart(formula = .outcome ~ ., control = list(minsplit = 20, minbucket = 7,
## cp = 0, maxcompete = 4, maxsurrogate = 5, usesurrogate = 2,
## surrogatestyle = 0, maxdepth = 30, xval = 0))
## n= 4475
##
## CP nsplit rel error
## 1 0.002225189 0 1
##
## Node number 1: 4475 observations
## predicted class=N expected loss=0.1673743 P(node) =1
## class counts: 3726 749
## probabilities: 0.833 0.167
##
## n= 4475
##
## node), split, n, loss, yval, (yprob)
## * denotes terminal node
##
## 1) root 4475 749 N (0.8326257 0.1673743) *
## [1] " calling mypredict_mdl for fit:"
## [1] "Classifier Probability Threshold: 0.5000 to maximize f.score.fit"
## Popular.fctr Popular.fctr.predict.Max.cor.Y.rpart.N
## 1 N 3726
## 2 Y 749
## Reference
## Prediction N Y
## N 3726 749
## Y 0 0
## Popular.fctr Popular.fctr.predict.Max.cor.Y.rpart.N
## 1 N 3726
## 2 Y 749
## Popular.fctr.predict.Max.cor.Y.rpart.Y
## 1 0
## 2 0
## [1] "in mypredict_mdl: calling confusionMatrix..."
## [1] " unique(df[, rsp_var_out]: "
## [1] N
## Levels: N Y
## [1] " unique(df[, rsp_var ]: "
## [1] Y N
## Levels: N Y
## Prediction
## Reference N Y
## N 3726 0
## Y 749 0
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 8.326257e-01 0.000000e+00 8.213602e-01 8.434553e-01 8.326257e-01
## AccuracyPValue McnemarPValue
## 5.097571e-01 1.800616e-164
## [1] " calling mypredict_mdl for OOB:"
## [1] "Classifier Probability Threshold: 0.5000 to maximize f.score.OOB"
## Popular.fctr Popular.fctr.predict.Max.cor.Y.rpart.N
## 1 N 1713
## 2 Y 344
## Reference
## Prediction N Y
## N 1713 344
## Y 0 0
## Popular.fctr Popular.fctr.predict.Max.cor.Y.rpart.N
## 1 N 1713
## 2 Y 344
## Popular.fctr.predict.Max.cor.Y.rpart.Y
## 1 0
## 2 0
## [1] "in mypredict_mdl: calling confusionMatrix..."
## [1] " unique(df[, rsp_var_out]: "
## [1] N
## Levels: N Y
## [1] " unique(df[, rsp_var ]: "
## [1] N Y
## Levels: N Y
## Prediction
## Reference N Y
## N 1713 0
## Y 344 0
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 8.327662e-01 0.000000e+00 8.159247e-01 8.486533e-01 8.327662e-01
## AccuracyPValue McnemarPValue
## 5.143944e-01 2.337097e-76
## model_id model_method feats max.nTuningRuns
## 1 Max.cor.Y.rpart rpart WordCount 3
## min.elapsedtime.everything min.elapsedtime.final max.auc.fit
## 1 1.265 0.063 0.5
## opt.prob.threshold.fit max.f.score.fit max.Accuracy.fit
## 1 0.5 0 0.8192183
## max.AccuracyLower.fit max.AccuracyUpper.fit max.Kappa.fit max.auc.OOB
## 1 0.8213602 0.8434553 0.0736708 0.5
## opt.prob.threshold.OOB max.f.score.OOB max.Accuracy.OOB
## 1 0.5 0 0.8327662
## max.AccuracyLower.OOB max.AccuracyUpper.OOB max.Kappa.OOB
## 1 0.8159247 0.8486533 0
## max.AccuracySD.fit max.KappaSD.fit
## 1 0.003061652 0.02524502# Used to compare vs. Interactions.High.cor.Y
ret_lst <- myfit_mdl(model_id="Max.cor.Y",
model_method=ifelse(glb_is_regression, "lm",
ifelse(glb_is_binomial, "glm", "rpart")),
model_type=glb_model_type,
indep_vars_vctr=max_cor_y_x_var,
rsp_var=glb_rsp_var, rsp_var_out=glb_rsp_var_out,
fit_df=glb_fitent_df, OOB_df=glb_OOBent_df,
n_cv_folds=glb_n_cv_folds, tune_models_df=NULL)## [1] "fitting model: Max.cor.Y.glm"
## [1] " indep_vars: WordCount"
## + Fold1: parameter=none
## - Fold1: parameter=none
## + Fold2: parameter=none
## - Fold2: parameter=none
## + Fold3: parameter=none
## - Fold3: parameter=none
## Aggregating results
## Fitting final model on full training set
##
## Call:
## NULL
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -2.8100 -0.5797 -0.4982 -0.4485 2.1935
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -2.318e+00 6.554e-02 -35.37 <2e-16 ***
## WordCount 1.202e-03 7.808e-05 15.39 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 4042.7 on 4474 degrees of freedom
## Residual deviance: 3774.8 on 4473 degrees of freedom
## AIC: 3778.8
##
## Number of Fisher Scoring iterations: 4
##
## [1] " calling mypredict_mdl for fit:"
## Reference
## Prediction N Y
## N 0 0
## Y 3726 749
## Popular.fctr Popular.fctr.predict.Max.cor.Y.glm.N
## 1 N 0
## 2 Y 0
## Popular.fctr.predict.Max.cor.Y.glm.Y
## 1 3726
## 2 749
## Reference
## Prediction N Y
## N 581 27
## Y 3145 722
## Popular.fctr Popular.fctr.predict.Max.cor.Y.glm.N
## 1 N 581
## 2 Y 27
## Popular.fctr.predict.Max.cor.Y.glm.Y
## 1 3145
## 2 722
## Reference
## Prediction N Y
## N 3088 378
## Y 638 371
## Popular.fctr Popular.fctr.predict.Max.cor.Y.glm.N
## 1 N 3088
## 2 Y 378
## Popular.fctr.predict.Max.cor.Y.glm.Y
## 1 638
## 2 371
## Reference
## Prediction N Y
## N 3515 628
## Y 211 121
## Popular.fctr Popular.fctr.predict.Max.cor.Y.glm.N
## 1 N 3515
## 2 Y 628
## Popular.fctr.predict.Max.cor.Y.glm.Y
## 1 211
## 2 121
## Reference
## Prediction N Y
## N 3625 702
## Y 101 47
## Popular.fctr Popular.fctr.predict.Max.cor.Y.glm.N
## 1 N 3625
## 2 Y 702
## Popular.fctr.predict.Max.cor.Y.glm.Y
## 1 101
## 2 47
## Reference
## Prediction N Y
## N 3659 727
## Y 67 22
## Popular.fctr Popular.fctr.predict.Max.cor.Y.glm.N
## 1 N 3659
## 2 Y 727
## Popular.fctr.predict.Max.cor.Y.glm.Y
## 1 67
## 2 22
## Reference
## Prediction N Y
## N 3693 734
## Y 33 15
## Popular.fctr Popular.fctr.predict.Max.cor.Y.glm.N
## 1 N 3693
## 2 Y 734
## Popular.fctr.predict.Max.cor.Y.glm.Y
## 1 33
## 2 15
## Reference
## Prediction N Y
## N 3708 740
## Y 18 9
## Popular.fctr Popular.fctr.predict.Max.cor.Y.glm.N
## 1 N 3708
## 2 Y 740
## Popular.fctr.predict.Max.cor.Y.glm.Y
## 1 18
## 2 9
## Reference
## Prediction N Y
## N 3717 742
## Y 9 7
## Popular.fctr Popular.fctr.predict.Max.cor.Y.glm.N
## 1 N 3717
## 2 Y 742
## Popular.fctr.predict.Max.cor.Y.glm.Y
## 1 9
## 2 7
## Reference
## Prediction N Y
## N 3724 743
## Y 2 6
## Popular.fctr Popular.fctr.predict.Max.cor.Y.glm.N
## 1 N 3724
## 2 Y 743
## Popular.fctr.predict.Max.cor.Y.glm.Y
## 1 2
## 2 6
## Reference
## Prediction N Y
## N 3726 749
## Y 0 0
## Popular.fctr Popular.fctr.predict.Max.cor.Y.glm.N
## 1 N 3726
## 2 Y 749
## Popular.fctr.predict.Max.cor.Y.glm.Y
## 1 0
## 2 0
## threshold f.score
## 1 0.0 0.28675345
## 2 0.1 0.31282496
## 3 0.2 0.42207053
## 4 0.3 0.22386679
## 5 0.4 0.10479376
## 6 0.5 0.05250597
## 7 0.6 0.03764115
## 8 0.7 0.02319588
## 9 0.8 0.01830065
## 10 0.9 0.01585205
## 11 1.0 0.00000000
## [1] "Classifier Probability Threshold: 0.2000 to maximize f.score.fit"
## Popular.fctr Popular.fctr.predict.Max.cor.Y.glm.N
## 1 N 3088
## 2 Y 378
## Popular.fctr.predict.Max.cor.Y.glm.Y
## 1 638
## 2 371
## Reference
## Prediction N Y
## N 3088 378
## Y 638 371
## Popular.fctr Popular.fctr.predict.Max.cor.Y.glm.N
## 1 N 3088
## 2 Y 378
## Popular.fctr.predict.Max.cor.Y.glm.Y
## 1 638
## 2 371
## [1] "in mypredict_mdl: calling confusionMatrix..."
## [1] " unique(df[, rsp_var_out]: "
## [1] N Y
## Levels: N Y
## [1] " unique(df[, rsp_var ]: "
## [1] Y N
## Levels: N Y
## Prediction
## Reference N Y
## N 3088 638
## Y 378 371
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 7.729609e-01 2.846273e-01 7.603993e-01 7.851652e-01 8.326257e-01
## AccuracyPValue McnemarPValue
## 1.000000e+00 4.453285e-16
## [1] " calling mypredict_mdl for OOB:"
## Reference
## Prediction N Y
## N 0 0
## Y 1713 344
## Popular.fctr Popular.fctr.predict.Max.cor.Y.glm.N
## 1 N 0
## 2 Y 0
## Popular.fctr.predict.Max.cor.Y.glm.Y
## 1 1713
## 2 344
## Reference
## Prediction N Y
## N 250 5
## Y 1463 339
## Popular.fctr Popular.fctr.predict.Max.cor.Y.glm.N
## 1 N 250
## 2 Y 5
## Popular.fctr.predict.Max.cor.Y.glm.Y
## 1 1463
## 2 339
## Reference
## Prediction N Y
## N 1418 182
## Y 295 162
## Popular.fctr Popular.fctr.predict.Max.cor.Y.glm.N
## 1 N 1418
## 2 Y 182
## Popular.fctr.predict.Max.cor.Y.glm.Y
## 1 295
## 2 162
## Reference
## Prediction N Y
## N 1602 272
## Y 111 72
## Popular.fctr Popular.fctr.predict.Max.cor.Y.glm.N
## 1 N 1602
## 2 Y 272
## Popular.fctr.predict.Max.cor.Y.glm.Y
## 1 111
## 2 72
## Reference
## Prediction N Y
## N 1653 314
## Y 60 30
## Popular.fctr Popular.fctr.predict.Max.cor.Y.glm.N
## 1 N 1653
## 2 Y 314
## Popular.fctr.predict.Max.cor.Y.glm.Y
## 1 60
## 2 30
## Reference
## Prediction N Y
## N 1670 329
## Y 43 15
## Popular.fctr Popular.fctr.predict.Max.cor.Y.glm.N
## 1 N 1670
## 2 Y 329
## Popular.fctr.predict.Max.cor.Y.glm.Y
## 1 43
## 2 15
## Reference
## Prediction N Y
## N 1686 336
## Y 27 8
## Popular.fctr Popular.fctr.predict.Max.cor.Y.glm.N
## 1 N 1686
## 2 Y 336
## Popular.fctr.predict.Max.cor.Y.glm.Y
## 1 27
## 2 8
## Reference
## Prediction N Y
## N 1700 341
## Y 13 3
## Popular.fctr Popular.fctr.predict.Max.cor.Y.glm.N
## 1 N 1700
## 2 Y 341
## Popular.fctr.predict.Max.cor.Y.glm.Y
## 1 13
## 2 3
## Reference
## Prediction N Y
## N 1708 341
## Y 5 3
## Popular.fctr Popular.fctr.predict.Max.cor.Y.glm.N
## 1 N 1708
## 2 Y 341
## Popular.fctr.predict.Max.cor.Y.glm.Y
## 1 5
## 2 3
## Reference
## Prediction N Y
## N 1710 342
## Y 3 2
## Popular.fctr Popular.fctr.predict.Max.cor.Y.glm.N
## 1 N 1710
## 2 Y 342
## Popular.fctr.predict.Max.cor.Y.glm.Y
## 1 3
## 2 2
## Reference
## Prediction N Y
## N 1713 344
## Y 0 0
## Popular.fctr Popular.fctr.predict.Max.cor.Y.glm.N
## 1 N 1713
## 2 Y 344
## Popular.fctr.predict.Max.cor.Y.glm.Y
## 1 0
## 2 0
## threshold f.score
## 1 0.0 0.28654727
## 2 0.1 0.31593663
## 3 0.2 0.40449438
## 4 0.3 0.27324478
## 5 0.4 0.13824885
## 6 0.5 0.07462687
## 7 0.6 0.04221636
## 8 0.7 0.01666667
## 9 0.8 0.01704545
## 10 0.9 0.01146132
## 11 1.0 0.00000000
## [1] "Classifier Probability Threshold: 0.2000 to maximize f.score.OOB"
## Popular.fctr Popular.fctr.predict.Max.cor.Y.glm.N
## 1 N 1418
## 2 Y 182
## Popular.fctr.predict.Max.cor.Y.glm.Y
## 1 295
## 2 162
## Reference
## Prediction N Y
## N 1418 182
## Y 295 162
## Popular.fctr Popular.fctr.predict.Max.cor.Y.glm.N
## 1 N 1418
## 2 Y 182
## Popular.fctr.predict.Max.cor.Y.glm.Y
## 1 295
## 2 162
## [1] "in mypredict_mdl: calling confusionMatrix..."
## [1] " unique(df[, rsp_var_out]: "
## [1] Y N
## Levels: N Y
## [1] " unique(df[, rsp_var ]: "
## [1] N Y
## Levels: N Y
## Prediction
## Reference N Y
## N 1418 295
## Y 182 162
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 7.681089e-01 2.640573e-01 7.492527e-01 7.861979e-01 8.327662e-01
## AccuracyPValue McnemarPValue
## 1.000000e+00 2.926379e-07
## model_id model_method feats max.nTuningRuns
## 1 Max.cor.Y.glm glm WordCount 1
## min.elapsedtime.everything min.elapsedtime.final max.auc.fit
## 1 1.126 0.062 0.7350785
## opt.prob.threshold.fit max.f.score.fit max.Accuracy.fit
## 1 0.2 0.4220705 0.8225702
## max.AccuracyLower.fit max.AccuracyUpper.fit max.Kappa.fit max.auc.OOB
## 1 0.7603993 0.7851652 0.01764536 0.7378291
## opt.prob.threshold.OOB max.f.score.OOB max.Accuracy.OOB
## 1 0.2 0.4044944 0.7681089
## max.AccuracyLower.OOB max.AccuracyUpper.OOB max.Kappa.OOB min.aic.fit
## 1 0.7492527 0.7861979 0.2640573 3778.81
## max.AccuracySD.fit max.KappaSD.fit
## 1 0.003663409 0.01234363# Interactions.High.cor.Y
if (length(int_feats <- setdiff(unique(glb_feats_df$cor.high.X), NA)) > 0) {
# lm & glm handle interaction terms; rpart & rf do not
if (glb_is_regression || glb_is_binomial) {
indep_vars_vctr <-
c(max_cor_y_x_var, paste(max_cor_y_x_var, int_feats, sep=":"))
} else { indep_vars_vctr <- union(max_cor_y_x_var, int_feats) }
ret_lst <- myfit_mdl(model_id="Interact.High.cor.Y",
model_method=ifelse(glb_is_regression, "lm",
ifelse(glb_is_binomial, "glm", "rpart")),
model_type=glb_model_type,
indep_vars_vctr,
glb_rsp_var, glb_rsp_var_out,
fit_df=glb_fitent_df, OOB_df=glb_OOBent_df,
n_cv_folds=glb_n_cv_folds, tune_models_df=NULL)
} ## [1] "fitting model: Interact.High.cor.Y.glm"
## [1] " indep_vars: WordCount, WordCount:S.year, WordCount:S.compani, WordCount:S.time, WordCount:S.will, WordCount:S.york, WordCount:S.new, WordCount:S.week, WordCount:H.num.chars, WordCount:A.num.chars, WordCount:H.num.words, WordCount:S.num.chars.log, WordCount:A.num.words, WordCount:H.num.chars.log, WordCount:S.num.words, WordCount:A.num.words.log, WordCount:S.num.chars"
## + Fold1: parameter=none
## - Fold1: parameter=none
## + Fold2: parameter=none
## - Fold2: parameter=none
## + Fold3: parameter=none
## - Fold3: parameter=none
## Aggregating results
## Fitting final model on full training set
##
## Call:
## NULL
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -3.9237 -0.5335 -0.4683 -0.4313 2.6081
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -2.352e+00 7.067e-02 -33.284 < 2e-16 ***
## WordCount 2.622e-02 2.948e-03 8.893 < 2e-16 ***
## `WordCount:S.year` -1.746e-04 2.874e-04 -0.607 0.543573
## `WordCount:S.compani` -3.102e-04 2.511e-04 -1.235 0.216652
## `WordCount:S.time` -4.821e-04 2.760e-04 -1.746 0.080739 .
## `WordCount:S.will` -2.414e-04 2.311e-04 -1.044 0.296267
## `WordCount:S.york` 3.580e-04 3.998e-04 0.895 0.370597
## `WordCount:S.new` -6.275e-05 1.931e-04 -0.325 0.745180
## `WordCount:S.week` -1.193e-03 3.820e-04 -3.123 0.001788 **
## `WordCount:H.num.chars` 2.976e-05 1.358e-05 2.192 0.028407 *
## `WordCount:A.num.chars` 4.853e-05 7.513e-05 0.646 0.518337
## `WordCount:H.num.words` -1.907e-04 6.670e-05 -2.859 0.004255 **
## `WordCount:S.num.chars.log` -7.817e-03 1.183e-03 -6.610 3.84e-11 ***
## `WordCount:A.num.words` -1.200e-03 7.341e-04 -1.635 0.102153
## `WordCount:H.num.chars.log` -5.536e-04 4.886e-04 -1.133 0.257175
## `WordCount:S.num.words` 7.942e-04 7.297e-04 1.088 0.276414
## `WordCount:A.num.words.log` 4.664e-03 1.248e-03 3.737 0.000186 ***
## `WordCount:S.num.chars` 1.000e-05 7.527e-05 0.133 0.894267
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 4042.7 on 4474 degrees of freedom
## Residual deviance: 3546.1 on 4457 degrees of freedom
## AIC: 3582.1
##
## Number of Fisher Scoring iterations: 6
##
## [1] " calling mypredict_mdl for fit:"
## Reference
## Prediction N Y
## N 0 0
## Y 3726 749
## Popular.fctr Popular.fctr.predict.Interact.High.cor.Y.glm.N
## 1 N 0
## 2 Y 0
## Popular.fctr.predict.Interact.High.cor.Y.glm.Y
## 1 3726
## 2 749
## Reference
## Prediction N Y
## N 1257 46
## Y 2469 703
## Popular.fctr Popular.fctr.predict.Interact.High.cor.Y.glm.N
## 1 N 1257
## 2 Y 46
## Popular.fctr.predict.Interact.High.cor.Y.glm.Y
## 1 2469
## 2 703
## Reference
## Prediction N Y
## N 3282 338
## Y 444 411
## Popular.fctr Popular.fctr.predict.Interact.High.cor.Y.glm.N
## 1 N 3282
## 2 Y 338
## Popular.fctr.predict.Interact.High.cor.Y.glm.Y
## 1 444
## 2 411
## Reference
## Prediction N Y
## N 3533 517
## Y 193 232
## Popular.fctr Popular.fctr.predict.Interact.High.cor.Y.glm.N
## 1 N 3533
## 2 Y 517
## Popular.fctr.predict.Interact.High.cor.Y.glm.Y
## 1 193
## 2 232
## Reference
## Prediction N Y
## N 3608 591
## Y 118 158
## Popular.fctr Popular.fctr.predict.Interact.High.cor.Y.glm.N
## 1 N 3608
## 2 Y 591
## Popular.fctr.predict.Interact.High.cor.Y.glm.Y
## 1 118
## 2 158
## Reference
## Prediction N Y
## N 3639 648
## Y 87 101
## Popular.fctr Popular.fctr.predict.Interact.High.cor.Y.glm.N
## 1 N 3639
## 2 Y 648
## Popular.fctr.predict.Interact.High.cor.Y.glm.Y
## 1 87
## 2 101
## Reference
## Prediction N Y
## N 3659 680
## Y 67 69
## Popular.fctr Popular.fctr.predict.Interact.High.cor.Y.glm.N
## 1 N 3659
## 2 Y 680
## Popular.fctr.predict.Interact.High.cor.Y.glm.Y
## 1 67
## 2 69
## Reference
## Prediction N Y
## N 3678 703
## Y 48 46
## Popular.fctr Popular.fctr.predict.Interact.High.cor.Y.glm.N
## 1 N 3678
## 2 Y 703
## Popular.fctr.predict.Interact.High.cor.Y.glm.Y
## 1 48
## 2 46
## Reference
## Prediction N Y
## N 3696 720
## Y 30 29
## Popular.fctr Popular.fctr.predict.Interact.High.cor.Y.glm.N
## 1 N 3696
## 2 Y 720
## Popular.fctr.predict.Interact.High.cor.Y.glm.Y
## 1 30
## 2 29
## Reference
## Prediction N Y
## N 3709 735
## Y 17 14
## Popular.fctr Popular.fctr.predict.Interact.High.cor.Y.glm.N
## 1 N 3709
## 2 Y 735
## Popular.fctr.predict.Interact.High.cor.Y.glm.Y
## 1 17
## 2 14
## Reference
## Prediction N Y
## N 3726 749
## Y 0 0
## Popular.fctr Popular.fctr.predict.Interact.High.cor.Y.glm.N
## 1 N 3726
## 2 Y 749
## Popular.fctr.predict.Interact.High.cor.Y.glm.Y
## 1 0
## 2 0
## threshold f.score
## 1 0.0 0.28675345
## 2 0.1 0.35858199
## 3 0.2 0.51246883
## 4 0.3 0.39522998
## 5 0.4 0.30829268
## 6 0.5 0.21558164
## 7 0.6 0.15593220
## 8 0.7 0.10913405
## 9 0.8 0.07178218
## 10 0.9 0.03589744
## 11 1.0 0.00000000
## [1] "Classifier Probability Threshold: 0.2000 to maximize f.score.fit"
## Popular.fctr Popular.fctr.predict.Interact.High.cor.Y.glm.N
## 1 N 3282
## 2 Y 338
## Popular.fctr.predict.Interact.High.cor.Y.glm.Y
## 1 444
## 2 411
## Reference
## Prediction N Y
## N 3282 338
## Y 444 411
## Popular.fctr Popular.fctr.predict.Interact.High.cor.Y.glm.N
## 1 N 3282
## 2 Y 338
## Popular.fctr.predict.Interact.High.cor.Y.glm.Y
## 1 444
## 2 411
## [1] "in mypredict_mdl: calling confusionMatrix..."
## [1] " unique(df[, rsp_var_out]: "
## [1] Y N
## Levels: N Y
## [1] " unique(df[, rsp_var ]: "
## [1] Y N
## Levels: N Y
## Prediction
## Reference N Y
## N 3282 444
## Y 338 411
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 0.8252513966 0.4065822072 0.8138021594 0.8362744678 0.8326256983
## AccuracyPValue McnemarPValue
## 0.9095557524 0.0001734853
## [1] " calling mypredict_mdl for OOB:"
## Reference
## Prediction N Y
## N 0 0
## Y 1713 344
## Popular.fctr Popular.fctr.predict.Interact.High.cor.Y.glm.N
## 1 N 0
## 2 Y 0
## Popular.fctr.predict.Interact.High.cor.Y.glm.Y
## 1 1713
## 2 344
## Reference
## Prediction N Y
## N 544 13
## Y 1169 331
## Popular.fctr Popular.fctr.predict.Interact.High.cor.Y.glm.N
## 1 N 544
## 2 Y 13
## Popular.fctr.predict.Interact.High.cor.Y.glm.Y
## 1 1169
## 2 331
## Reference
## Prediction N Y
## N 1484 153
## Y 229 191
## Popular.fctr Popular.fctr.predict.Interact.High.cor.Y.glm.N
## 1 N 1484
## 2 Y 153
## Popular.fctr.predict.Interact.High.cor.Y.glm.Y
## 1 229
## 2 191
## Reference
## Prediction N Y
## N 1607 237
## Y 106 107
## Popular.fctr Popular.fctr.predict.Interact.High.cor.Y.glm.N
## 1 N 1607
## 2 Y 237
## Popular.fctr.predict.Interact.High.cor.Y.glm.Y
## 1 106
## 2 107
## Reference
## Prediction N Y
## N 1649 273
## Y 64 71
## Popular.fctr Popular.fctr.predict.Interact.High.cor.Y.glm.N
## 1 N 1649
## 2 Y 273
## Popular.fctr.predict.Interact.High.cor.Y.glm.Y
## 1 64
## 2 71
## Reference
## Prediction N Y
## N 1669 294
## Y 44 50
## Popular.fctr Popular.fctr.predict.Interact.High.cor.Y.glm.N
## 1 N 1669
## 2 Y 294
## Popular.fctr.predict.Interact.High.cor.Y.glm.Y
## 1 44
## 2 50
## Reference
## Prediction N Y
## N 1685 311
## Y 28 33
## Popular.fctr Popular.fctr.predict.Interact.High.cor.Y.glm.N
## 1 N 1685
## 2 Y 311
## Popular.fctr.predict.Interact.High.cor.Y.glm.Y
## 1 28
## 2 33
## Reference
## Prediction N Y
## N 1690 327
## Y 23 17
## Popular.fctr Popular.fctr.predict.Interact.High.cor.Y.glm.N
## 1 N 1690
## 2 Y 327
## Popular.fctr.predict.Interact.High.cor.Y.glm.Y
## 1 23
## 2 17
## Reference
## Prediction N Y
## N 1698 336
## Y 15 8
## Popular.fctr Popular.fctr.predict.Interact.High.cor.Y.glm.N
## 1 N 1698
## 2 Y 336
## Popular.fctr.predict.Interact.High.cor.Y.glm.Y
## 1 15
## 2 8
## Reference
## Prediction N Y
## N 1704 339
## Y 9 5
## Popular.fctr Popular.fctr.predict.Interact.High.cor.Y.glm.N
## 1 N 1704
## 2 Y 339
## Popular.fctr.predict.Interact.High.cor.Y.glm.Y
## 1 9
## 2 5
## Reference
## Prediction N Y
## N 1713 344
## Y 0 0
## Popular.fctr Popular.fctr.predict.Interact.High.cor.Y.glm.N
## 1 N 1713
## 2 Y 344
## Popular.fctr.predict.Interact.High.cor.Y.glm.Y
## 1 0
## 2 0
## threshold f.score
## 1 0.0 0.28654727
## 2 0.1 0.35900217
## 3 0.2 0.50000000
## 4 0.3 0.38420108
## 5 0.4 0.29645094
## 6 0.5 0.22831050
## 7 0.6 0.16296296
## 8 0.7 0.08854167
## 9 0.8 0.04359673
## 10 0.9 0.02793296
## 11 1.0 0.00000000
## [1] "Classifier Probability Threshold: 0.2000 to maximize f.score.OOB"
## Popular.fctr Popular.fctr.predict.Interact.High.cor.Y.glm.N
## 1 N 1484
## 2 Y 153
## Popular.fctr.predict.Interact.High.cor.Y.glm.Y
## 1 229
## 2 191
## Reference
## Prediction N Y
## N 1484 153
## Y 229 191
## Popular.fctr Popular.fctr.predict.Interact.High.cor.Y.glm.N
## 1 N 1484
## 2 Y 153
## Popular.fctr.predict.Interact.High.cor.Y.glm.Y
## 1 229
## 2 191
## [1] "in mypredict_mdl: calling confusionMatrix..."
## [1] " unique(df[, rsp_var_out]: "
## [1] N Y
## Levels: N Y
## [1] " unique(df[, rsp_var ]: "
## [1] N Y
## Levels: N Y
## Prediction
## Reference N Y
## N 1484 229
## Y 153 191
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 0.8142926592 0.3873527586 0.7968011314 0.8308835268 0.8327661643
## AccuracyPValue McnemarPValue
## 0.9877368727 0.0001243773
## model_id model_method
## 1 Interact.High.cor.Y.glm glm
## feats
## 1 WordCount, WordCount:S.year, WordCount:S.compani, WordCount:S.time, WordCount:S.will, WordCount:S.york, WordCount:S.new, WordCount:S.week, WordCount:H.num.chars, WordCount:A.num.chars, WordCount:H.num.words, WordCount:S.num.chars.log, WordCount:A.num.words, WordCount:H.num.chars.log, WordCount:S.num.words, WordCount:A.num.words.log, WordCount:S.num.chars
## max.nTuningRuns min.elapsedtime.everything min.elapsedtime.final
## 1 1 1.834 0.278
## max.auc.fit opt.prob.threshold.fit max.f.score.fit max.Accuracy.fit
## 1 0.8062116 0.2 0.5124688 0.8359773
## max.AccuracyLower.fit max.AccuracyUpper.fit max.Kappa.fit max.auc.OOB
## 1 0.8138022 0.8362745 0.1618606 0.815698
## opt.prob.threshold.OOB max.f.score.OOB max.Accuracy.OOB
## 1 0.2 0.5 0.8142927
## max.AccuracyLower.OOB max.AccuracyUpper.OOB max.Kappa.OOB min.aic.fit
## 1 0.7968011 0.8308835 0.3873528 3582.118
## max.AccuracySD.fit max.KappaSD.fit
## 1 0.003949821 0.0313674# Low.cor.X
if (glb_is_classification && glb_is_binomial)
indep_vars_vctr <- subset(glb_feats_df, is.na(cor.high.X) &
is.ConditionalX.y &
(exclude.as.feat != 1))[, "id"] else
indep_vars_vctr <- subset(glb_feats_df, is.na(cor.high.X) &
(exclude.as.feat != 1))[, "id"]
ret_lst <- myfit_mdl(model_id="Low.cor.X",
model_method=ifelse(glb_is_regression, "lm",
ifelse(glb_is_binomial, "glm", "rpart")),
indep_vars_vctr=indep_vars_vctr,
model_type=glb_model_type,
glb_rsp_var, glb_rsp_var_out,
fit_df=glb_fitent_df, OOB_df=glb_OOBent_df,
n_cv_folds=glb_n_cv_folds, tune_models_df=NULL)## [1] "fitting model: Low.cor.X.glm"
## [1] " indep_vars: WordCount, PubDate.hour, PubDate.month, .rnorm, PubDate.date, PubDate.second, PubDate.minute, PubDate.wkday, S.year, S.compani, S.time, S.will, S.york, S.new, H.new, S.week, SectionName.fctr, NewsDesk.fctr, SubsectionName.fctr, H.num.chars, A.num.chars, H.num.words, H.num.words.log, S.num.chars.log, A.num.words, A.num.words.unq, A.num.words.log, A.num.words.unq.log"
## + Fold1: parameter=none## Warning in predict.lm(object, newdata, se.fit, scale = 1, type =
## ifelse(type == : prediction from a rank-deficient fit may be misleading## - Fold1: parameter=none
## + Fold2: parameter=none## Warning in predict.lm(object, newdata, se.fit, scale = 1, type =
## ifelse(type == : prediction from a rank-deficient fit may be misleading## - Fold2: parameter=none
## + Fold3: parameter=none## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred## Warning in predict.lm(object, newdata, se.fit, scale = 1, type =
## ifelse(type == : prediction from a rank-deficient fit may be misleading## - Fold3: parameter=none
## Aggregating results
## Fitting final model on full training set## Warning: not plotting observations with leverage one:
## 297
## Warning: not plotting observations with leverage one:
## 297
##
## Call:
## NULL
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -3.2760 -0.3642 -0.2129 -0.0002 3.4410
##
## Coefficients: (7 not defined because of singularities)
## Estimate Std. Error z value
## (Intercept) 3.176e+00 1.446e+00 2.196
## WordCount 1.709e-03 1.368e-04 12.488
## PubDate.hour 2.717e-02 1.250e-02 2.173
## PubDate.month -1.049e-01 7.316e-02 -1.434
## .rnorm -3.150e-02 5.706e-02 -0.552
## PubDate.date -2.904e-03 6.729e-03 -0.432
## PubDate.second -7.826e-04 3.330e-03 -0.235
## PubDate.minute -3.938e-03 3.256e-03 -1.209
## PubDate.wkday -1.250e-01 3.677e-02 -3.401
## S.year -3.537e-01 2.942e-01 -1.202
## S.compani -2.970e-01 2.710e-01 -1.096
## S.time -4.330e-01 2.471e-01 -1.752
## S.will -2.559e-01 2.121e-01 -1.206
## S.york 2.782e-01 3.905e-01 0.712
## S.new -1.126e-01 1.978e-01 -0.569
## H.new -6.285e-01 3.261e-01 -1.927
## S.week -2.703e-01 2.499e-01 -1.082
## SectionName.fctrArts -3.497e+00 3.925e-01 -8.910
## `SectionName.fctrBusiness Day` -5.018e+00 7.034e-01 -7.133
## SectionName.fctrHealth -7.987e-01 1.785e+00 -0.447
## SectionName.fctrOpinion -1.333e+00 1.538e+00 -0.866
## SectionName.fctr -2.193e+00 1.042e+00 -2.104
## SectionName.fctrStyle -2.869e+00 4.647e+03 -0.001
## SectionName.fctrWorld -3.737e+00 3.262e+03 -0.001
## SectionName.fctrTechnology -2.572e+00 4.199e-01 -6.126
## SectionName.fctrMagazine -1.982e+01 1.337e+03 -0.015
## SectionName.fctrMultimedia -4.481e+00 1.453e+00 -3.085
## SectionName.fctrTravel -4.856e+00 1.062e+00 -4.573
## SectionName.fctrU.S. 1.653e+01 7.397e+02 0.022
## `SectionName.fctrN.Y. / Region` -2.876e+00 4.786e-01 -6.008
## SectionName.fctrOpen -1.870e+01 4.590e+03 -0.004
## SectionName.fctrSports -2.049e+01 6.523e+03 -0.003
## NewsDesk.fctrCulture NA NA NA
## NewsDesk.fctrScience 3.011e-01 1.758e+00 0.171
## NewsDesk.fctrOpEd 1.227e+00 1.519e+00 0.808
## NewsDesk.fctr -1.191e+00 1.001e+00 -1.190
## NewsDesk.fctrForeign -1.678e+01 6.006e+02 -0.028
## NewsDesk.fctrStyles -1.727e+01 7.397e+02 -0.023
## NewsDesk.fctrTStyle -3.457e+00 1.079e+00 -3.203
## NewsDesk.fctrMagazine NA NA NA
## NewsDesk.fctrTravel NA NA NA
## NewsDesk.fctrMetro NA NA NA
## NewsDesk.fctrNational -1.676e+01 4.567e+03 -0.004
## NewsDesk.fctrSports NA NA NA
## SubsectionName.fctrDealbook 1.407e+00 6.289e-01 2.238
## `SubsectionName.fctrRoom For Debate` -6.279e+00 1.556e+00 -4.036
## `SubsectionName.fctrFashion & Style` NA NA NA
## `SubsectionName.fctrAsia Pacific` 1.459e+01 3.206e+03 0.005
## SubsectionName.fctrEducation -3.487e+01 8.487e+02 -0.041
## `SubsectionName.fctrThe Public Editor` 1.622e+00 1.287e+00 1.260
## `SubsectionName.fctrSmall Business` NA NA NA
## SubsectionName.fctrPolitics -1.872e+01 6.528e+03 -0.003
## H.num.chars 2.105e-02 7.977e-03 2.639
## A.num.chars 8.331e-03 5.538e-03 1.504
## H.num.words -7.865e-02 1.466e-01 -0.536
## H.num.words.log -6.812e-01 7.494e-01 -0.909
## S.num.chars.log 2.360e-01 5.663e-01 0.417
## A.num.words -4.139e-01 3.020e-01 -1.371
## A.num.words.unq 3.317e-01 3.142e-01 1.056
## A.num.words.log 5.155e+00 4.437e+00 1.162
## A.num.words.unq.log -6.047e+00 4.513e+00 -1.340
## Pr(>|z|)
## (Intercept) 0.028083 *
## WordCount < 2e-16 ***
## PubDate.hour 0.029802 *
## PubDate.month 0.151599
## .rnorm 0.580927
## PubDate.date 0.666049
## PubDate.second 0.814202
## PubDate.minute 0.226549
## PubDate.wkday 0.000672 ***
## S.year 0.229205
## S.compani 0.273123
## S.time 0.079706 .
## S.will 0.227638
## S.york 0.476199
## S.new 0.569182
## H.new 0.053929 .
## S.week 0.279370
## SectionName.fctrArts < 2e-16 ***
## `SectionName.fctrBusiness Day` 9.81e-13 ***
## SectionName.fctrHealth 0.654623
## SectionName.fctrOpinion 0.386223
## SectionName.fctr 0.035340 *
## SectionName.fctrStyle 0.999507
## SectionName.fctrWorld 0.999086
## SectionName.fctrTechnology 9.03e-10 ***
## SectionName.fctrMagazine 0.988171
## SectionName.fctrMultimedia 0.002035 **
## SectionName.fctrTravel 4.80e-06 ***
## SectionName.fctrU.S. 0.982174
## `SectionName.fctrN.Y. / Region` 1.87e-09 ***
## SectionName.fctrOpen 0.996749
## SectionName.fctrSports 0.997493
## NewsDesk.fctrCulture NA
## NewsDesk.fctrScience 0.864000
## NewsDesk.fctrOpEd 0.419183
## NewsDesk.fctr 0.233972
## NewsDesk.fctrForeign 0.977710
## NewsDesk.fctrStyles 0.981375
## NewsDesk.fctrTStyle 0.001360 **
## NewsDesk.fctrMagazine NA
## NewsDesk.fctrTravel NA
## NewsDesk.fctrMetro NA
## NewsDesk.fctrNational 0.997071
## NewsDesk.fctrSports NA
## SubsectionName.fctrDealbook 0.025252 *
## `SubsectionName.fctrRoom For Debate` 5.43e-05 ***
## `SubsectionName.fctrFashion & Style` NA
## `SubsectionName.fctrAsia Pacific` 0.996370
## SubsectionName.fctrEducation 0.967226
## `SubsectionName.fctrThe Public Editor` 0.207530
## `SubsectionName.fctrSmall Business` NA
## SubsectionName.fctrPolitics 0.997712
## H.num.chars 0.008316 **
## A.num.chars 0.132493
## H.num.words 0.591692
## H.num.words.log 0.363367
## S.num.chars.log 0.676855
## A.num.words 0.170403
## A.num.words.unq 0.291142
## A.num.words.log 0.245223
## A.num.words.unq.log 0.180266
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 4042.7 on 4474 degrees of freedom
## Residual deviance: 2098.3 on 4421 degrees of freedom
## AIC: 2206.3
##
## Number of Fisher Scoring iterations: 17
##
## [1] " calling mypredict_mdl for fit:"## Warning in predict.lm(object, newdata, se.fit, scale = 1, type =
## ifelse(type == : prediction from a rank-deficient fit may be misleading## Warning in predict.lm(object, newdata, se.fit, scale = 1, type =
## ifelse(type == : prediction from a rank-deficient fit may be misleading
## Reference
## Prediction N Y
## N 0 0
## Y 3726 749
## Popular.fctr Popular.fctr.predict.Low.cor.X.glm.N
## 1 N 0
## 2 Y 0
## Popular.fctr.predict.Low.cor.X.glm.Y
## 1 3726
## 2 749
## Reference
## Prediction N Y
## N 3037 57
## Y 689 692
## Popular.fctr Popular.fctr.predict.Low.cor.X.glm.N
## 1 N 3037
## 2 Y 57
## Popular.fctr.predict.Low.cor.X.glm.Y
## 1 689
## 2 692
## Reference
## Prediction N Y
## N 3362 145
## Y 364 604
## Popular.fctr Popular.fctr.predict.Low.cor.X.glm.N
## 1 N 3362
## 2 Y 145
## Popular.fctr.predict.Low.cor.X.glm.Y
## 1 364
## 2 604
## Reference
## Prediction N Y
## N 3483 192
## Y 243 557
## Popular.fctr Popular.fctr.predict.Low.cor.X.glm.N
## 1 N 3483
## 2 Y 192
## Popular.fctr.predict.Low.cor.X.glm.Y
## 1 243
## 2 557
## Reference
## Prediction N Y
## N 3528 222
## Y 198 527
## Popular.fctr Popular.fctr.predict.Low.cor.X.glm.N
## 1 N 3528
## 2 Y 222
## Popular.fctr.predict.Low.cor.X.glm.Y
## 1 198
## 2 527
## Reference
## Prediction N Y
## N 3572 266
## Y 154 483
## Popular.fctr Popular.fctr.predict.Low.cor.X.glm.N
## 1 N 3572
## 2 Y 266
## Popular.fctr.predict.Low.cor.X.glm.Y
## 1 154
## 2 483
## Reference
## Prediction N Y
## N 3607 309
## Y 119 440
## Popular.fctr Popular.fctr.predict.Low.cor.X.glm.N
## 1 N 3607
## 2 Y 309
## Popular.fctr.predict.Low.cor.X.glm.Y
## 1 119
## 2 440
## Reference
## Prediction N Y
## N 3645 385
## Y 81 364
## Popular.fctr Popular.fctr.predict.Low.cor.X.glm.N
## 1 N 3645
## 2 Y 385
## Popular.fctr.predict.Low.cor.X.glm.Y
## 1 81
## 2 364
## Reference
## Prediction N Y
## N 3687 503
## Y 39 246
## Popular.fctr Popular.fctr.predict.Low.cor.X.glm.N
## 1 N 3687
## 2 Y 503
## Popular.fctr.predict.Low.cor.X.glm.Y
## 1 39
## 2 246
## Reference
## Prediction N Y
## N 3700 644
## Y 26 105
## Popular.fctr Popular.fctr.predict.Low.cor.X.glm.N
## 1 N 3700
## 2 Y 644
## Popular.fctr.predict.Low.cor.X.glm.Y
## 1 26
## 2 105
## Reference
## Prediction N Y
## N 3726 749
## Y 0 0
## Popular.fctr Popular.fctr.predict.Low.cor.X.glm.N
## 1 N 3726
## 2 Y 749
## Popular.fctr.predict.Low.cor.X.glm.Y
## 1 0
## 2 0
## threshold f.score
## 1 0.0 0.2867534
## 2 0.1 0.6497653
## 3 0.2 0.7035527
## 4 0.3 0.7191737
## 5 0.4 0.7150611
## 6 0.5 0.6969697
## 7 0.6 0.6727829
## 8 0.7 0.6097152
## 9 0.8 0.4758221
## 10 0.9 0.2386364
## 11 1.0 0.0000000## [1] "Classifier Probability Threshold: 0.3000 to maximize f.score.fit"
## Popular.fctr Popular.fctr.predict.Low.cor.X.glm.N
## 1 N 3483
## 2 Y 192
## Popular.fctr.predict.Low.cor.X.glm.Y
## 1 243
## 2 557
## Reference
## Prediction N Y
## N 3483 192
## Y 243 557
## Popular.fctr Popular.fctr.predict.Low.cor.X.glm.N
## 1 N 3483
## 2 Y 192
## Popular.fctr.predict.Low.cor.X.glm.Y
## 1 243
## 2 557
## [1] "in mypredict_mdl: calling confusionMatrix..."
## [1] " unique(df[, rsp_var_out]: "
## [1] Y N
## Levels: N Y
## [1] " unique(df[, rsp_var ]: "
## [1] Y N
## Levels: N Y
## Prediction
## Reference N Y
## N 3483 243
## Y 192 557
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 9.027933e-01 6.604749e-01 8.937373e-01 9.113196e-01 8.326257e-01
## AccuracyPValue McnemarPValue
## 2.150408e-41 1.651565e-02
## [1] " calling mypredict_mdl for OOB:"## Warning in predict.lm(object, newdata, se.fit, scale = 1, type =
## ifelse(type == : prediction from a rank-deficient fit may be misleading## Warning in predict.lm(object, newdata, se.fit, scale = 1, type =
## ifelse(type == : prediction from a rank-deficient fit may be misleading
## Reference
## Prediction N Y
## N 0 0
## Y 1713 344
## Popular.fctr Popular.fctr.predict.Low.cor.X.glm.N
## 1 N 0
## 2 Y 0
## Popular.fctr.predict.Low.cor.X.glm.Y
## 1 1713
## 2 344
## Reference
## Prediction N Y
## N 1379 26
## Y 334 318
## Popular.fctr Popular.fctr.predict.Low.cor.X.glm.N
## 1 N 1379
## 2 Y 26
## Popular.fctr.predict.Low.cor.X.glm.Y
## 1 334
## 2 318
## Reference
## Prediction N Y
## N 1533 59
## Y 180 285
## Popular.fctr Popular.fctr.predict.Low.cor.X.glm.N
## 1 N 1533
## 2 Y 59
## Popular.fctr.predict.Low.cor.X.glm.Y
## 1 180
## 2 285
## Reference
## Prediction N Y
## N 1592 86
## Y 121 258
## Popular.fctr Popular.fctr.predict.Low.cor.X.glm.N
## 1 N 1592
## 2 Y 86
## Popular.fctr.predict.Low.cor.X.glm.Y
## 1 121
## 2 258
## Reference
## Prediction N Y
## N 1615 102
## Y 98 242
## Popular.fctr Popular.fctr.predict.Low.cor.X.glm.N
## 1 N 1615
## 2 Y 102
## Popular.fctr.predict.Low.cor.X.glm.Y
## 1 98
## 2 242
## Reference
## Prediction N Y
## N 1635 111
## Y 78 233
## Popular.fctr Popular.fctr.predict.Low.cor.X.glm.N
## 1 N 1635
## 2 Y 111
## Popular.fctr.predict.Low.cor.X.glm.Y
## 1 78
## 2 233
## Reference
## Prediction N Y
## N 1645 132
## Y 68 212
## Popular.fctr Popular.fctr.predict.Low.cor.X.glm.N
## 1 N 1645
## 2 Y 132
## Popular.fctr.predict.Low.cor.X.glm.Y
## 1 68
## 2 212
## Reference
## Prediction N Y
## N 1671 164
## Y 42 180
## Popular.fctr Popular.fctr.predict.Low.cor.X.glm.N
## 1 N 1671
## 2 Y 164
## Popular.fctr.predict.Low.cor.X.glm.Y
## 1 42
## 2 180
## Reference
## Prediction N Y
## N 1688 232
## Y 25 112
## Popular.fctr Popular.fctr.predict.Low.cor.X.glm.N
## 1 N 1688
## 2 Y 232
## Popular.fctr.predict.Low.cor.X.glm.Y
## 1 25
## 2 112
## Reference
## Prediction N Y
## N 1702 284
## Y 11 60
## Popular.fctr Popular.fctr.predict.Low.cor.X.glm.N
## 1 N 1702
## 2 Y 284
## Popular.fctr.predict.Low.cor.X.glm.Y
## 1 11
## 2 60
## Reference
## Prediction N Y
## N 1713 344
## Y 0 0
## Popular.fctr Popular.fctr.predict.Low.cor.X.glm.N
## 1 N 1713
## 2 Y 344
## Popular.fctr.predict.Low.cor.X.glm.Y
## 1 0
## 2 0
## threshold f.score
## 1 0.0 0.2865473
## 2 0.1 0.6385542
## 3 0.2 0.7045735
## 4 0.3 0.7136929
## 5 0.4 0.7076023
## 6 0.5 0.7114504
## 7 0.6 0.6794872
## 8 0.7 0.6360424
## 9 0.8 0.4656965
## 10 0.9 0.2891566
## 11 1.0 0.0000000
## [1] "Classifier Probability Threshold: 0.3000 to maximize f.score.OOB"
## Popular.fctr Popular.fctr.predict.Low.cor.X.glm.N
## 1 N 1592
## 2 Y 86
## Popular.fctr.predict.Low.cor.X.glm.Y
## 1 121
## 2 258
## Reference
## Prediction N Y
## N 1592 86
## Y 121 258
## Popular.fctr Popular.fctr.predict.Low.cor.X.glm.N
## 1 N 1592
## 2 Y 86
## Popular.fctr.predict.Low.cor.X.glm.Y
## 1 121
## 2 258
## [1] "in mypredict_mdl: calling confusionMatrix..."
## [1] " unique(df[, rsp_var_out]: "
## [1] N Y
## Levels: N Y
## [1] " unique(df[, rsp_var ]: "
## [1] N Y
## Levels: N Y
## Prediction
## Reference N Y
## N 1592 121
## Y 86 258
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 8.993680e-01 6.528225e-01 8.855497e-01 9.120357e-01 8.327662e-01
## AccuracyPValue McnemarPValue
## 4.579965e-18 1.811968e-02
## model_id model_method
## 1 Low.cor.X.glm glm
## feats
## 1 WordCount, PubDate.hour, PubDate.month, .rnorm, PubDate.date, PubDate.second, PubDate.minute, PubDate.wkday, S.year, S.compani, S.time, S.will, S.york, S.new, H.new, S.week, SectionName.fctr, NewsDesk.fctr, SubsectionName.fctr, H.num.chars, A.num.chars, H.num.words, H.num.words.log, S.num.chars.log, A.num.words, A.num.words.unq, A.num.words.log, A.num.words.unq.log
## max.nTuningRuns min.elapsedtime.everything min.elapsedtime.final
## 1 1 4.791 1.252
## max.auc.fit opt.prob.threshold.fit max.f.score.fit max.Accuracy.fit
## 1 0.9372948 0.3 0.7191737 0.9003357
## max.AccuracyLower.fit max.AccuracyUpper.fit max.Kappa.fit max.auc.OOB
## 1 0.8937373 0.9113196 0.619684 0.9326067
## opt.prob.threshold.OOB max.f.score.OOB max.Accuracy.OOB
## 1 0.3 0.7136929 0.899368
## max.AccuracyLower.OOB max.AccuracyUpper.OOB max.Kappa.OOB min.aic.fit
## 1 0.8855497 0.9120357 0.6528225 2206.319
## max.AccuracySD.fit max.KappaSD.fit
## 1 0.002679281 0.006873523glb_script_df <- rbind(glb_script_df,
data.frame(chunk_label="fit.models",
chunk_step_major=glb_script_df[nrow(glb_script_df), "chunk_step_major"],
chunk_step_minor=glb_script_df[nrow(glb_script_df), "chunk_step_minor"]+1,
elapsed=(proc.time() - glb_script_tm)["elapsed"]))
print(tail(glb_script_df, 2))## chunk_label chunk_step_major chunk_step_minor elapsed
## elapsed9 fit.models 6 0 140.736
## elapsed10 fit.models 6 1 180.884# All X that is not user excluded
if (glb_is_classification && glb_is_binomial) {
model_id_pfx <- "Conditional.X"
# indep_vars_vctr <- setdiff(names(glb_fitent_df), union(glb_rsp_var, glb_exclude_vars_as_features))
indep_vars_vctr <- subset(glb_feats_df, is.ConditionalX.y &
(exclude.as.feat != 1))[, "id"]
} else {
model_id_pfx <- "All.X"
indep_vars_vctr <- subset(glb_feats_df,
(exclude.as.feat != 1))[, "id"]
}
for (method in glb_models_method_vctr) {
ret_lst <- myfit_mdl(model_id=paste0(model_id_pfx, ""), model_method=method,
indep_vars_vctr=indep_vars_vctr,
model_type=glb_model_type,
rsp_var=glb_rsp_var, rsp_var_out=glb_rsp_var_out,
fit_df=glb_fitent_df, OOB_df=glb_OOBent_df,
n_cv_folds=glb_n_cv_folds, tune_models_df=glb_tune_models_df)
# Since caret does not optimize rpart well
if (method == "rpart")
ret_lst <- myfit_mdl(model_id=paste0(model_id_pfx, ".cp.0"), model_method=method,
indep_vars_vctr=indep_vars_vctr,
model_type=glb_model_type,
rsp_var=glb_rsp_var, rsp_var_out=glb_rsp_var_out,
fit_df=glb_fitent_df, OOB_df=glb_OOBent_df,
n_cv_folds=0, tune_models_df=data.frame(parameter="cp", min=0.0, max=0.0, by=0.1))
# Compare how rf performs w/i & w/o .rnorm
if (method == "rf")
ret_lst <- myfit_mdl(model_id=paste0(model_id_pfx, ".no.rnorm"), model_method=method,
indep_vars_vctr=setdiff(indep_vars_vctr, c(".rnorm")),
model_type=glb_model_type,
rsp_var=glb_rsp_var, rsp_var_out=glb_rsp_var_out,
fit_df=glb_fitent_df, OOB_df=glb_OOBent_df,
n_cv_folds=glb_n_cv_folds, tune_models_df=glb_tune_models_df)
}## [1] "fitting model: Conditional.X.glm"
## [1] " indep_vars: WordCount, PubDate.hour, PubDate.month, .rnorm, PubDate.date, PubDate.second, PubDate.minute, PubDate.wkday, S.year, A.year, S.compani, A.compani, S.time, A.time, S.will, A.will, S.york, A.york, S.new, A.new, H.new, S.week, A.week, SectionName.fctr, NewsDesk.fctr, SubsectionName.fctr, H.num.chars, H.num.chars.log, A.num.chars, S.num.chars, H.num.words, H.num.words.unq, H.num.words.log, S.num.chars.log, A.num.chars.log, A.num.words, S.num.words, H.num.words.unq.log, A.num.words.unq, S.num.words.unq, A.num.words.log, S.num.words.log, A.num.words.unq.log, S.num.words.unq.log"
## + Fold1: parameter=none## Warning in predict.lm(object, newdata, se.fit, scale = 1, type =
## ifelse(type == : prediction from a rank-deficient fit may be misleading## - Fold1: parameter=none
## + Fold2: parameter=none## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred## Warning in predict.lm(object, newdata, se.fit, scale = 1, type =
## ifelse(type == : prediction from a rank-deficient fit may be misleading## - Fold2: parameter=none
## + Fold3: parameter=none## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred## Warning in predict.lm(object, newdata, se.fit, scale = 1, type =
## ifelse(type == : prediction from a rank-deficient fit may be misleading## - Fold3: parameter=none
## Aggregating results
## Fitting final model on full training set## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred## Warning: not plotting observations with leverage one:
## 297, 4229, 4338
## Warning: not plotting observations with leverage one:
## 297, 4229, 4338
## Warning in sqrt(crit * p * (1 - hh)/hh): NaNs produced## Warning in sqrt(crit * p * (1 - hh)/hh): NaNs produced##
## Call:
## NULL
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -3.3481 -0.3588 -0.2092 -0.0001 3.4697
##
## Coefficients: (10 not defined because of singularities)
## Estimate Std. Error z value
## (Intercept) 6.406e+00 2.002e+00 3.200
## WordCount 1.723e-03 1.381e-04 12.475
## PubDate.hour 2.821e-02 1.254e-02 2.250
## PubDate.month -9.863e-02 7.364e-02 -1.339
## .rnorm -3.235e-02 5.749e-02 -0.563
## PubDate.date -2.133e-03 6.773e-03 -0.315
## PubDate.second -8.456e-04 3.347e-03 -0.253
## PubDate.minute -3.413e-03 3.289e-03 -1.038
## PubDate.wkday -1.282e-01 3.706e-02 -3.459
## S.year -4.140e-01 2.991e-01 -1.384
## A.year NA NA NA
## S.compani -7.379e-01 1.186e+04 0.000
## A.compani 4.521e-01 1.186e+04 0.000
## S.time 2.207e+01 1.075e+04 0.002
## A.time -2.252e+01 1.075e+04 -0.002
## S.will 7.264e-02 5.001e+03 0.000
## A.will -3.097e-01 5.001e+03 0.000
## S.york 3.019e-01 3.920e-01 0.770
## A.york NA NA NA
## S.new 7.876e+00 6.839e+03 0.001
## A.new -7.970e+00 6.839e+03 -0.001
## H.new -6.222e-01 3.212e-01 -1.937
## S.week -2.468e-01 2.500e-01 -0.987
## A.week NA NA NA
## SectionName.fctrArts -3.447e+00 3.992e-01 -8.636
## `SectionName.fctrBusiness Day` -5.063e+00 7.101e-01 -7.130
## SectionName.fctrHealth -1.001e+00 1.792e+00 -0.559
## SectionName.fctrOpinion -1.336e+00 1.543e+00 -0.865
## SectionName.fctr -2.284e+00 1.047e+00 -2.181
## SectionName.fctrStyle -3.025e+00 7.571e+03 0.000
## SectionName.fctrWorld -3.298e+00 4.572e+03 -0.001
## SectionName.fctrTechnology -2.574e+00 4.286e-01 -6.005
## SectionName.fctrMagazine -2.085e+01 2.207e+03 -0.009
## SectionName.fctrMultimedia -4.500e+00 1.458e+00 -3.087
## SectionName.fctrTravel -4.803e+00 1.064e+00 -4.514
## SectionName.fctrU.S. 1.744e+01 1.204e+03 0.014
## `SectionName.fctrN.Y. / Region` -2.845e+00 4.820e-01 -5.903
## SectionName.fctrOpen -1.986e+01 7.581e+03 -0.003
## SectionName.fctrSports -2.150e+01 1.075e+04 -0.002
## NewsDesk.fctrCulture NA NA NA
## NewsDesk.fctrScience 5.345e-01 1.762e+00 0.303
## NewsDesk.fctrOpEd 1.309e+00 1.522e+00 0.860
## NewsDesk.fctr -1.117e+00 1.004e+00 -1.112
## NewsDesk.fctrForeign -1.762e+01 9.903e+02 -0.018
## NewsDesk.fctrStyles -1.817e+01 1.204e+03 -0.015
## NewsDesk.fctrTStyle -3.420e+00 1.084e+00 -3.157
## NewsDesk.fctrMagazine NA NA NA
## NewsDesk.fctrTravel NA NA NA
## NewsDesk.fctrMetro NA NA NA
## NewsDesk.fctrNational -1.766e+01 7.539e+03 -0.002
## NewsDesk.fctrSports NA NA NA
## SubsectionName.fctrDealbook 1.446e+00 6.310e-01 2.291
## `SubsectionName.fctrRoom For Debate` -6.342e+00 1.558e+00 -4.070
## `SubsectionName.fctrFashion & Style` NA NA NA
## `SubsectionName.fctrAsia Pacific` 1.487e+01 4.464e+03 0.003
## SubsectionName.fctrEducation -4.828e+01 1.265e+03 -0.038
## `SubsectionName.fctrThe Public Editor` 1.602e+00 1.289e+00 1.243
## `SubsectionName.fctrSmall Business` NA NA NA
## SubsectionName.fctrPolitics -1.972e+01 1.077e+04 -0.002
## H.num.chars 5.119e-02 2.025e-02 2.528
## H.num.chars.log -1.341e+00 8.375e-01 -1.601
## A.num.chars 2.096e-01 1.849e-01 1.133
## S.num.chars -1.955e-01 1.851e-01 -1.056
## H.num.words 6.568e-01 1.053e+00 0.623
## H.num.words.unq -9.759e-01 1.046e+00 -0.933
## H.num.words.log -1.838e+00 6.144e+00 -0.299
## S.num.chars.log 5.826e+01 5.541e+01 1.052
## A.num.chars.log -5.840e+01 5.540e+01 -1.054
## A.num.words 1.257e+00 4.442e+00 0.283
## S.num.words -1.731e+00 4.476e+00 -0.387
## H.num.words.unq.log 2.559e+00 6.076e+00 0.421
## A.num.words.unq -7.306e+00 5.197e+00 -1.406
## S.num.words.unq 7.678e+00 5.236e+00 1.466
## A.num.words.log -3.348e+01 1.080e+02 -0.310
## S.num.words.log 3.890e+01 1.083e+02 0.359
## A.num.words.unq.log 1.661e+02 1.165e+02 1.425
## S.num.words.unq.log -1.722e+02 1.169e+02 -1.473
## Pr(>|z|)
## (Intercept) 0.001375 **
## WordCount < 2e-16 ***
## PubDate.hour 0.024461 *
## PubDate.month 0.180443
## .rnorm 0.573608
## PubDate.date 0.752824
## PubDate.second 0.800519
## PubDate.minute 0.299431
## PubDate.wkday 0.000543 ***
## S.year 0.166335
## A.year NA
## S.compani 0.999950
## A.compani 0.999970
## S.time 0.998363
## A.time 0.998329
## S.will 0.999988
## A.will 0.999951
## S.york 0.441208
## A.york NA
## S.new 0.999081
## A.new 0.999070
## H.new 0.052758 .
## S.week 0.323592
## A.week NA
## SectionName.fctrArts < 2e-16 ***
## `SectionName.fctrBusiness Day` 1.01e-12 ***
## SectionName.fctrHealth 0.576397
## SectionName.fctrOpinion 0.386848
## SectionName.fctr 0.029194 *
## SectionName.fctrStyle 0.999681
## SectionName.fctrWorld 0.999425
## SectionName.fctrTechnology 1.92e-09 ***
## SectionName.fctrMagazine 0.992463
## SectionName.fctrMultimedia 0.002020 **
## SectionName.fctrTravel 6.35e-06 ***
## SectionName.fctrU.S. 0.988439
## `SectionName.fctrN.Y. / Region` 3.58e-09 ***
## SectionName.fctrOpen 0.997909
## SectionName.fctrSports 0.998405
## NewsDesk.fctrCulture NA
## NewsDesk.fctrScience 0.761699
## NewsDesk.fctrOpEd 0.389789
## NewsDesk.fctr 0.266009
## NewsDesk.fctrForeign 0.985806
## NewsDesk.fctrStyles 0.987956
## NewsDesk.fctrTStyle 0.001596 **
## NewsDesk.fctrMagazine NA
## NewsDesk.fctrTravel NA
## NewsDesk.fctrMetro NA
## NewsDesk.fctrNational 0.998131
## NewsDesk.fctrSports NA
## SubsectionName.fctrDealbook 0.021967 *
## `SubsectionName.fctrRoom For Debate` 4.70e-05 ***
## `SubsectionName.fctrFashion & Style` NA
## `SubsectionName.fctrAsia Pacific` 0.997342
## SubsectionName.fctrEducation 0.969566
## `SubsectionName.fctrThe Public Editor` 0.213863
## `SubsectionName.fctrSmall Business` NA
## SubsectionName.fctrPolitics 0.998539
## H.num.chars 0.011472 *
## H.num.chars.log 0.109308
## A.num.chars 0.257005
## S.num.chars 0.290872
## H.num.words 0.532967
## H.num.words.unq 0.351029
## H.num.words.log 0.764811
## S.num.chars.log 0.293020
## A.num.chars.log 0.291740
## A.num.words 0.777187
## S.num.words 0.698987
## H.num.words.unq.log 0.673625
## A.num.words.unq 0.159753
## S.num.words.unq 0.142594
## A.num.words.log 0.756572
## S.num.words.log 0.719513
## A.num.words.unq.log 0.154158
## S.num.words.unq.log 0.140835
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 4042.7 on 4474 degrees of freedom
## Residual deviance: 2078.3 on 4408 degrees of freedom
## AIC: 2212.3
##
## Number of Fisher Scoring iterations: 18
##
## [1] " calling mypredict_mdl for fit:"## Warning in predict.lm(object, newdata, se.fit, scale = 1, type =
## ifelse(type == : prediction from a rank-deficient fit may be misleading## Warning in predict.lm(object, newdata, se.fit, scale = 1, type =
## ifelse(type == : prediction from a rank-deficient fit may be misleading
## Reference
## Prediction N Y
## N 0 0
## Y 3726 749
## Popular.fctr Popular.fctr.predict.Conditional.X.glm.N
## 1 N 0
## 2 Y 0
## Popular.fctr.predict.Conditional.X.glm.Y
## 1 3726
## 2 749
## Reference
## Prediction N Y
## N 3046 59
## Y 680 690
## Popular.fctr Popular.fctr.predict.Conditional.X.glm.N
## 1 N 3046
## 2 Y 59
## Popular.fctr.predict.Conditional.X.glm.Y
## 1 680
## 2 690
## Reference
## Prediction N Y
## N 3362 140
## Y 364 609
## Popular.fctr Popular.fctr.predict.Conditional.X.glm.N
## 1 N 3362
## 2 Y 140
## Popular.fctr.predict.Conditional.X.glm.Y
## 1 364
## 2 609
## Reference
## Prediction N Y
## N 3481 190
## Y 245 559
## Popular.fctr Popular.fctr.predict.Conditional.X.glm.N
## 1 N 3481
## 2 Y 190
## Popular.fctr.predict.Conditional.X.glm.Y
## 1 245
## 2 559
## Reference
## Prediction N Y
## N 3528 225
## Y 198 524
## Popular.fctr Popular.fctr.predict.Conditional.X.glm.N
## 1 N 3528
## 2 Y 225
## Popular.fctr.predict.Conditional.X.glm.Y
## 1 198
## 2 524
## Reference
## Prediction N Y
## N 3571 262
## Y 155 487
## Popular.fctr Popular.fctr.predict.Conditional.X.glm.N
## 1 N 3571
## 2 Y 262
## Popular.fctr.predict.Conditional.X.glm.Y
## 1 155
## 2 487
## Reference
## Prediction N Y
## N 3610 310
## Y 116 439
## Popular.fctr Popular.fctr.predict.Conditional.X.glm.N
## 1 N 3610
## 2 Y 310
## Popular.fctr.predict.Conditional.X.glm.Y
## 1 116
## 2 439
## Reference
## Prediction N Y
## N 3651 383
## Y 75 366
## Popular.fctr Popular.fctr.predict.Conditional.X.glm.N
## 1 N 3651
## 2 Y 383
## Popular.fctr.predict.Conditional.X.glm.Y
## 1 75
## 2 366
## Reference
## Prediction N Y
## N 3687 499
## Y 39 250
## Popular.fctr Popular.fctr.predict.Conditional.X.glm.N
## 1 N 3687
## 2 Y 499
## Popular.fctr.predict.Conditional.X.glm.Y
## 1 39
## 2 250
## Reference
## Prediction N Y
## N 3701 635
## Y 25 114
## Popular.fctr Popular.fctr.predict.Conditional.X.glm.N
## 1 N 3701
## 2 Y 635
## Popular.fctr.predict.Conditional.X.glm.Y
## 1 25
## 2 114
## Reference
## Prediction N Y
## N 3726 749
## Y 0 0
## Popular.fctr Popular.fctr.predict.Conditional.X.glm.N
## 1 N 3726
## 2 Y 749
## Popular.fctr.predict.Conditional.X.glm.Y
## 1 0
## 2 0
## threshold f.score
## 1 0.0 0.2867534
## 2 0.1 0.6512506
## 3 0.2 0.7073171
## 4 0.3 0.7198970
## 5 0.4 0.7124405
## 6 0.5 0.7002157
## 7 0.6 0.6733129
## 8 0.7 0.6151261
## 9 0.8 0.4816956
## 10 0.9 0.2567568
## 11 1.0 0.0000000## [1] "Classifier Probability Threshold: 0.3000 to maximize f.score.fit"
## Popular.fctr Popular.fctr.predict.Conditional.X.glm.N
## 1 N 3481
## 2 Y 190
## Popular.fctr.predict.Conditional.X.glm.Y
## 1 245
## 2 559
## Reference
## Prediction N Y
## N 3481 190
## Y 245 559
## Popular.fctr Popular.fctr.predict.Conditional.X.glm.N
## 1 N 3481
## 2 Y 190
## Popular.fctr.predict.Conditional.X.glm.Y
## 1 245
## 2 559
## [1] "in mypredict_mdl: calling confusionMatrix..."
## [1] " unique(df[, rsp_var_out]: "
## [1] Y N
## Levels: N Y
## [1] " unique(df[, rsp_var ]: "
## [1] Y N
## Levels: N Y
## Prediction
## Reference N Y
## N 3481 245
## Y 190 559
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 9.027933e-01 6.611786e-01 8.937373e-01 9.113196e-01 8.326257e-01
## AccuracyPValue McnemarPValue
## 2.150408e-41 9.622663e-03
## [1] " calling mypredict_mdl for OOB:"## Warning in predict.lm(object, newdata, se.fit, scale = 1, type =
## ifelse(type == : prediction from a rank-deficient fit may be misleading## Warning in predict.lm(object, newdata, se.fit, scale = 1, type =
## ifelse(type == : prediction from a rank-deficient fit may be misleading
## Reference
## Prediction N Y
## N 0 0
## Y 1713 344
## Popular.fctr Popular.fctr.predict.Conditional.X.glm.N
## 1 N 0
## 2 Y 0
## Popular.fctr.predict.Conditional.X.glm.Y
## 1 1713
## 2 344
## Reference
## Prediction N Y
## N 1388 27
## Y 325 317
## Popular.fctr Popular.fctr.predict.Conditional.X.glm.N
## 1 N 1388
## 2 Y 27
## Popular.fctr.predict.Conditional.X.glm.Y
## 1 325
## 2 317
## Reference
## Prediction N Y
## N 1530 59
## Y 183 285
## Popular.fctr Popular.fctr.predict.Conditional.X.glm.N
## 1 N 1530
## 2 Y 59
## Popular.fctr.predict.Conditional.X.glm.Y
## 1 183
## 2 285
## Reference
## Prediction N Y
## N 1590 88
## Y 123 256
## Popular.fctr Popular.fctr.predict.Conditional.X.glm.N
## 1 N 1590
## 2 Y 88
## Popular.fctr.predict.Conditional.X.glm.Y
## 1 123
## 2 256
## Reference
## Prediction N Y
## N 1613 101
## Y 100 243
## Popular.fctr Popular.fctr.predict.Conditional.X.glm.N
## 1 N 1613
## 2 Y 101
## Popular.fctr.predict.Conditional.X.glm.Y
## 1 100
## 2 243
## Reference
## Prediction N Y
## N 1633 112
## Y 80 232
## Popular.fctr Popular.fctr.predict.Conditional.X.glm.N
## 1 N 1633
## 2 Y 112
## Popular.fctr.predict.Conditional.X.glm.Y
## 1 80
## 2 232
## Reference
## Prediction N Y
## N 1646 127
## Y 67 217
## Popular.fctr Popular.fctr.predict.Conditional.X.glm.N
## 1 N 1646
## 2 Y 127
## Popular.fctr.predict.Conditional.X.glm.Y
## 1 67
## 2 217
## Reference
## Prediction N Y
## N 1671 159
## Y 42 185
## Popular.fctr Popular.fctr.predict.Conditional.X.glm.N
## 1 N 1671
## 2 Y 159
## Popular.fctr.predict.Conditional.X.glm.Y
## 1 42
## 2 185
## Reference
## Prediction N Y
## N 1687 221
## Y 26 123
## Popular.fctr Popular.fctr.predict.Conditional.X.glm.N
## 1 N 1687
## 2 Y 221
## Popular.fctr.predict.Conditional.X.glm.Y
## 1 26
## 2 123
## Reference
## Prediction N Y
## N 1701 285
## Y 12 59
## Popular.fctr Popular.fctr.predict.Conditional.X.glm.N
## 1 N 1701
## 2 Y 285
## Popular.fctr.predict.Conditional.X.glm.Y
## 1 12
## 2 59
## Reference
## Prediction N Y
## N 1713 344
## Y 0 0
## Popular.fctr Popular.fctr.predict.Conditional.X.glm.N
## 1 N 1713
## 2 Y 344
## Popular.fctr.predict.Conditional.X.glm.Y
## 1 0
## 2 0
## threshold f.score
## 1 0.0 0.2865473
## 2 0.1 0.6430020
## 3 0.2 0.7019704
## 4 0.3 0.7081604
## 5 0.4 0.7074236
## 6 0.5 0.7073171
## 7 0.6 0.6910828
## 8 0.7 0.6479860
## 9 0.8 0.4989858
## 10 0.9 0.2843373
## 11 1.0 0.0000000
## [1] "Classifier Probability Threshold: 0.3000 to maximize f.score.OOB"
## Popular.fctr Popular.fctr.predict.Conditional.X.glm.N
## 1 N 1590
## 2 Y 88
## Popular.fctr.predict.Conditional.X.glm.Y
## 1 123
## 2 256
## Reference
## Prediction N Y
## N 1590 88
## Y 123 256
## Popular.fctr Popular.fctr.predict.Conditional.X.glm.N
## 1 N 1590
## 2 Y 88
## Popular.fctr.predict.Conditional.X.glm.Y
## 1 123
## 2 256
## [1] "in mypredict_mdl: calling confusionMatrix..."
## [1] " unique(df[, rsp_var_out]: "
## [1] N Y
## Levels: N Y
## [1] " unique(df[, rsp_var ]: "
## [1] N Y
## Levels: N Y
## Prediction
## Reference N Y
## N 1590 123
## Y 88 256
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 8.974234e-01 6.461137e-01 8.834974e-01 9.102047e-01 8.327662e-01
## AccuracyPValue McnemarPValue
## 4.637200e-17 1.924987e-02
## model_id model_method
## 1 Conditional.X.glm glm
## feats
## 1 WordCount, PubDate.hour, PubDate.month, .rnorm, PubDate.date, PubDate.second, PubDate.minute, PubDate.wkday, S.year, A.year, S.compani, A.compani, S.time, A.time, S.will, A.will, S.york, A.york, S.new, A.new, H.new, S.week, A.week, SectionName.fctr, NewsDesk.fctr, SubsectionName.fctr, H.num.chars, H.num.chars.log, A.num.chars, S.num.chars, H.num.words, H.num.words.unq, H.num.words.log, S.num.chars.log, A.num.chars.log, A.num.words, S.num.words, H.num.words.unq.log, A.num.words.unq, S.num.words.unq, A.num.words.log, S.num.words.log, A.num.words.unq.log, S.num.words.unq.log
## max.nTuningRuns min.elapsedtime.everything min.elapsedtime.final
## 1 1 6.841 1.746
## max.auc.fit opt.prob.threshold.fit max.f.score.fit max.Accuracy.fit
## 1 0.9381279 0.3 0.719897 0.8992181
## max.AccuracyLower.fit max.AccuracyUpper.fit max.Kappa.fit max.auc.OOB
## 1 0.8937373 0.9113196 0.6173657 0.9327051
## opt.prob.threshold.OOB max.f.score.OOB max.Accuracy.OOB
## 1 0.3 0.7081604 0.8974234
## max.AccuracyLower.OOB max.AccuracyUpper.OOB max.Kappa.OOB min.aic.fit
## 1 0.8834974 0.9102047 0.6461137 2212.301
## max.AccuracySD.fit max.KappaSD.fit
## 1 0.002140938 0.00177553
## [1] "fitting model: Conditional.X.rpart"
## [1] " indep_vars: WordCount, PubDate.hour, PubDate.month, PubDate.date, PubDate.second, PubDate.minute, PubDate.wkday, S.year, A.year, S.compani, A.compani, S.time, A.time, S.will, A.will, S.york, A.york, S.new, A.new, H.new, S.week, A.week, SectionName.fctr, NewsDesk.fctr, SubsectionName.fctr, H.num.chars, H.num.chars.log, A.num.chars, S.num.chars, H.num.words, H.num.words.unq, H.num.words.log, S.num.chars.log, A.num.chars.log, A.num.words, S.num.words, H.num.words.unq.log, A.num.words.unq, S.num.words.unq, A.num.words.log, S.num.words.log, A.num.words.unq.log, S.num.words.unq.log"
## + Fold1: cp=0.02804
## - Fold1: cp=0.02804
## + Fold2: cp=0.02804
## - Fold2: cp=0.02804
## + Fold3: cp=0.02804
## - Fold3: cp=0.02804
## Aggregating results
## Selecting tuning parameters
## Fitting cp = 0.028 on full training set## Warning in myfit_mdl(model_id = paste0(model_id_pfx, ""), model_method =
## method, : model's bestTune found at an extreme of tuneGrid for parameter:
## cp
## Call:
## rpart(formula = .outcome ~ ., control = list(minsplit = 20, minbucket = 7,
## cp = 0, maxcompete = 4, maxsurrogate = 5, usesurrogate = 2,
## surrogatestyle = 0, maxdepth = 30, xval = 0))
## n= 4475
##
## CP nsplit rel error
## 1 0.25901202 0 1.0000000
## 2 0.03938585 1 0.7409880
## 3 0.02803738 3 0.6622163
##
## Variable importance
## NewsDesk.fctrOpEd SectionName.fctrOpinion WordCount
## 32 26 12
## SectionName.fctrHealth NewsDesk.fctrScience A.num.chars
## 6 6 4
## A.num.chars.log S.num.chars S.num.chars.log
## 4 4 4
## PubDate.hour
## 1
##
## Node number 1: 4475 observations, complexity param=0.259012
## predicted class=N expected loss=0.1673743 P(node) =1
## class counts: 3726 749
## probabilities: 0.833 0.167
## left son=2 (4131 obs) right son=3 (344 obs)
## Primary splits:
## NewsDesk.fctrOpEd < 0.5 to the left, improve=281.52360, (0 missing)
## SectionName.fctrOpinion < 0.5 to the left, improve=244.96950, (0 missing)
## WordCount < 683.5 to the left, improve=113.02630, (0 missing)
## A.num.chars < 44.5 to the right, improve= 93.81257, (0 missing)
## S.num.chars < 44.5 to the right, improve= 93.81257, (0 missing)
## Surrogate splits:
## SectionName.fctrOpinion < 0.5 to the left, agree=0.986, adj=0.817, (0 split)
## A.num.chars < 40.5 to the right, agree=0.933, adj=0.128, (0 split)
## S.num.chars < 40.5 to the right, agree=0.933, adj=0.128, (0 split)
## S.num.chars.log < 3.725621 to the right, agree=0.933, adj=0.128, (0 split)
## A.num.chars.log < 3.725621 to the right, agree=0.933, adj=0.128, (0 split)
##
## Node number 2: 4131 observations, complexity param=0.03938585
## predicted class=N expected loss=0.1161946 P(node) =0.9231285
## class counts: 3651 480
## probabilities: 0.884 0.116
## left son=4 (3066 obs) right son=5 (1065 obs)
## Primary splits:
## WordCount < 683.5 to the left, improve=105.55990, (0 missing)
## NewsDesk.fctrScience < 0.5 to the left, improve= 70.50963, (0 missing)
## SectionName.fctrHealth < 0.5 to the left, improve= 69.73481, (0 missing)
## PubDate.hour < 21.5 to the left, improve= 53.45620, (0 missing)
## NewsDesk.fctrStyles < 0.5 to the left, improve= 27.64076, (0 missing)
## Surrogate splits:
## SectionName.fctrOpinion < 0.5 to the left, agree=0.755, adj=0.048, (0 split)
## SubsectionName.fctrRoom For Debate < 0.5 to the left, agree=0.753, adj=0.041, (0 split)
## SectionName.fctrHealth < 0.5 to the left, agree=0.745, adj=0.012, (0 split)
## NewsDesk.fctrScience < 0.5 to the left, agree=0.745, adj=0.011, (0 split)
## SubsectionName.fctrThe Public Editor < 0.5 to the left, agree=0.744, adj=0.008, (0 split)
##
## Node number 3: 344 observations
## predicted class=Y expected loss=0.2180233 P(node) =0.07687151
## class counts: 75 269
## probabilities: 0.218 0.782
##
## Node number 4: 3066 observations
## predicted class=N expected loss=0.04957599 P(node) =0.6851397
## class counts: 2914 152
## probabilities: 0.950 0.050
##
## Node number 5: 1065 observations, complexity param=0.03938585
## predicted class=N expected loss=0.3079812 P(node) =0.2379888
## class counts: 737 328
## probabilities: 0.692 0.308
## left son=10 (994 obs) right son=11 (71 obs)
## Primary splits:
## SectionName.fctrHealth < 0.5 to the left, improve=56.15144, (0 missing)
## NewsDesk.fctrScience < 0.5 to the left, improve=55.08535, (0 missing)
## PubDate.hour < 7.5 to the left, improve=18.57852, (0 missing)
## NewsDesk.fctrForeign < 0.5 to the right, improve=11.91201, (0 missing)
## SectionName.fctrWorld < 0.5 to the right, improve=11.48588, (0 missing)
## Surrogate splits:
## NewsDesk.fctrScience < 0.5 to the left, agree=0.999, adj=0.986, (0 split)
## PubDate.hour < 0.5 to the right, agree=0.939, adj=0.085, (0 split)
##
## Node number 10: 994 observations
## predicted class=N expected loss=0.2645875 P(node) =0.2221229
## class counts: 731 263
## probabilities: 0.735 0.265
##
## Node number 11: 71 observations
## predicted class=Y expected loss=0.08450704 P(node) =0.01586592
## class counts: 6 65
## probabilities: 0.085 0.915
##
## n= 4475
##
## node), split, n, loss, yval, (yprob)
## * denotes terminal node
##
## 1) root 4475 749 N (0.83262570 0.16737430)
## 2) NewsDesk.fctrOpEd< 0.5 4131 480 N (0.88380537 0.11619463)
## 4) WordCount< 683.5 3066 152 N (0.95042401 0.04957599) *
## 5) WordCount>=683.5 1065 328 N (0.69201878 0.30798122)
## 10) SectionName.fctrHealth< 0.5 994 263 N (0.73541247 0.26458753) *
## 11) SectionName.fctrHealth>=0.5 71 6 Y (0.08450704 0.91549296) *
## 3) NewsDesk.fctrOpEd>=0.5 344 75 Y (0.21802326 0.78197674) *
## [1] " calling mypredict_mdl for fit:"
## Reference
## Prediction N Y
## N 0 0
## Y 3726 749
## Popular.fctr Popular.fctr.predict.Conditional.X.rpart.N
## 1 N 0
## 2 Y 0
## Popular.fctr.predict.Conditional.X.rpart.Y
## 1 3726
## 2 749
## Reference
## Prediction N Y
## N 2914 152
## Y 812 597
## Popular.fctr Popular.fctr.predict.Conditional.X.rpart.N
## 1 N 2914
## 2 Y 152
## Popular.fctr.predict.Conditional.X.rpart.Y
## 1 812
## 2 597
## Reference
## Prediction N Y
## N 2914 152
## Y 812 597
## Popular.fctr Popular.fctr.predict.Conditional.X.rpart.N
## 1 N 2914
## 2 Y 152
## Popular.fctr.predict.Conditional.X.rpart.Y
## 1 812
## 2 597
## Reference
## Prediction N Y
## N 3645 415
## Y 81 334
## Popular.fctr Popular.fctr.predict.Conditional.X.rpart.N
## 1 N 3645
## 2 Y 415
## Popular.fctr.predict.Conditional.X.rpart.Y
## 1 81
## 2 334
## Reference
## Prediction N Y
## N 3645 415
## Y 81 334
## Popular.fctr Popular.fctr.predict.Conditional.X.rpart.N
## 1 N 3645
## 2 Y 415
## Popular.fctr.predict.Conditional.X.rpart.Y
## 1 81
## 2 334
## Reference
## Prediction N Y
## N 3645 415
## Y 81 334
## Popular.fctr Popular.fctr.predict.Conditional.X.rpart.N
## 1 N 3645
## 2 Y 415
## Popular.fctr.predict.Conditional.X.rpart.Y
## 1 81
## 2 334
## Reference
## Prediction N Y
## N 3645 415
## Y 81 334
## Popular.fctr Popular.fctr.predict.Conditional.X.rpart.N
## 1 N 3645
## 2 Y 415
## Popular.fctr.predict.Conditional.X.rpart.Y
## 1 81
## 2 334
## Reference
## Prediction N Y
## N 3645 415
## Y 81 334
## Popular.fctr Popular.fctr.predict.Conditional.X.rpart.N
## 1 N 3645
## 2 Y 415
## Popular.fctr.predict.Conditional.X.rpart.Y
## 1 81
## 2 334
## Reference
## Prediction N Y
## N 3720 684
## Y 6 65
## Popular.fctr Popular.fctr.predict.Conditional.X.rpart.N
## 1 N 3720
## 2 Y 684
## Popular.fctr.predict.Conditional.X.rpart.Y
## 1 6
## 2 65
## Reference
## Prediction N Y
## N 3720 684
## Y 6 65
## Popular.fctr Popular.fctr.predict.Conditional.X.rpart.N
## 1 N 3720
## 2 Y 684
## Popular.fctr.predict.Conditional.X.rpart.Y
## 1 6
## 2 65
## Reference
## Prediction N Y
## N 3726 749
## Y 0 0
## Popular.fctr Popular.fctr.predict.Conditional.X.rpart.N
## 1 N 3726
## 2 Y 749
## Popular.fctr.predict.Conditional.X.rpart.Y
## 1 0
## 2 0
## threshold f.score
## 1 0.0 0.2867534
## 2 0.1 0.5532901
## 3 0.2 0.5532901
## 4 0.3 0.5738832
## 5 0.4 0.5738832
## 6 0.5 0.5738832
## 7 0.6 0.5738832
## 8 0.7 0.5738832
## 9 0.8 0.1585366
## 10 0.9 0.1585366
## 11 1.0 0.0000000
## [1] "Classifier Probability Threshold: 0.7000 to maximize f.score.fit"
## Popular.fctr Popular.fctr.predict.Conditional.X.rpart.N
## 1 N 3645
## 2 Y 415
## Popular.fctr.predict.Conditional.X.rpart.Y
## 1 81
## 2 334
## Reference
## Prediction N Y
## N 3645 415
## Y 81 334
## Popular.fctr Popular.fctr.predict.Conditional.X.rpart.N
## 1 N 3645
## 2 Y 415
## Popular.fctr.predict.Conditional.X.rpart.Y
## 1 81
## 2 334
## [1] "in mypredict_mdl: calling confusionMatrix..."
## [1] " unique(df[, rsp_var_out]: "
## [1] N Y
## Levels: N Y
## [1] " unique(df[, rsp_var ]: "
## [1] Y N
## Levels: N Y
## Prediction
## Reference N Y
## N 3645 81
## Y 415 334
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 8.891620e-01 5.161350e-01 8.795973e-01 8.982154e-01 8.326257e-01
## AccuracyPValue McnemarPValue
## 9.017296e-27 1.508203e-50
## [1] " calling mypredict_mdl for OOB:"
## Reference
## Prediction N Y
## N 0 0
## Y 1713 344
## Popular.fctr Popular.fctr.predict.Conditional.X.rpart.N
## 1 N 0
## 2 Y 0
## Popular.fctr.predict.Conditional.X.rpart.Y
## 1 1713
## 2 344
## Reference
## Prediction N Y
## N 1311 74
## Y 402 270
## Popular.fctr Popular.fctr.predict.Conditional.X.rpart.N
## 1 N 1311
## 2 Y 74
## Popular.fctr.predict.Conditional.X.rpart.Y
## 1 402
## 2 270
## Reference
## Prediction N Y
## N 1311 74
## Y 402 270
## Popular.fctr Popular.fctr.predict.Conditional.X.rpart.N
## 1 N 1311
## 2 Y 74
## Popular.fctr.predict.Conditional.X.rpart.Y
## 1 402
## 2 270
## Reference
## Prediction N Y
## N 1671 178
## Y 42 166
## Popular.fctr Popular.fctr.predict.Conditional.X.rpart.N
## 1 N 1671
## 2 Y 178
## Popular.fctr.predict.Conditional.X.rpart.Y
## 1 42
## 2 166
## Reference
## Prediction N Y
## N 1671 178
## Y 42 166
## Popular.fctr Popular.fctr.predict.Conditional.X.rpart.N
## 1 N 1671
## 2 Y 178
## Popular.fctr.predict.Conditional.X.rpart.Y
## 1 42
## 2 166
## Reference
## Prediction N Y
## N 1671 178
## Y 42 166
## Popular.fctr Popular.fctr.predict.Conditional.X.rpart.N
## 1 N 1671
## 2 Y 178
## Popular.fctr.predict.Conditional.X.rpart.Y
## 1 42
## 2 166
## Reference
## Prediction N Y
## N 1671 178
## Y 42 166
## Popular.fctr Popular.fctr.predict.Conditional.X.rpart.N
## 1 N 1671
## 2 Y 178
## Popular.fctr.predict.Conditional.X.rpart.Y
## 1 42
## 2 166
## Reference
## Prediction N Y
## N 1671 178
## Y 42 166
## Popular.fctr Popular.fctr.predict.Conditional.X.rpart.N
## 1 N 1671
## 2 Y 178
## Popular.fctr.predict.Conditional.X.rpart.Y
## 1 42
## 2 166
## Reference
## Prediction N Y
## N 1709 317
## Y 4 27
## Popular.fctr Popular.fctr.predict.Conditional.X.rpart.N
## 1 N 1709
## 2 Y 317
## Popular.fctr.predict.Conditional.X.rpart.Y
## 1 4
## 2 27
## Reference
## Prediction N Y
## N 1709 317
## Y 4 27
## Popular.fctr Popular.fctr.predict.Conditional.X.rpart.N
## 1 N 1709
## 2 Y 317
## Popular.fctr.predict.Conditional.X.rpart.Y
## 1 4
## 2 27
## Reference
## Prediction N Y
## N 1713 344
## Y 0 0
## Popular.fctr Popular.fctr.predict.Conditional.X.rpart.N
## 1 N 1713
## 2 Y 344
## Popular.fctr.predict.Conditional.X.rpart.Y
## 1 0
## 2 0
## threshold f.score
## 1 0.0 0.2865473
## 2 0.1 0.5314961
## 3 0.2 0.5314961
## 4 0.3 0.6014493
## 5 0.4 0.6014493
## 6 0.5 0.6014493
## 7 0.6 0.6014493
## 8 0.7 0.6014493
## 9 0.8 0.1440000
## 10 0.9 0.1440000
## 11 1.0 0.0000000
## [1] "Classifier Probability Threshold: 0.7000 to maximize f.score.OOB"
## Popular.fctr Popular.fctr.predict.Conditional.X.rpart.N
## 1 N 1671
## 2 Y 178
## Popular.fctr.predict.Conditional.X.rpart.Y
## 1 42
## 2 166
## Reference
## Prediction N Y
## N 1671 178
## Y 42 166
## Popular.fctr Popular.fctr.predict.Conditional.X.rpart.N
## 1 N 1671
## 2 Y 178
## Popular.fctr.predict.Conditional.X.rpart.Y
## 1 42
## 2 166
## [1] "in mypredict_mdl: calling confusionMatrix..."
## [1] " unique(df[, rsp_var_out]: "
## [1] N Y
## Levels: N Y
## [1] " unique(df[, rsp_var ]: "
## [1] N Y
## Levels: N Y
## Prediction
## Reference N Y
## N 1671 42
## Y 178 166
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 8.930481e-01 5.439760e-01 8.788850e-01 9.060795e-01 8.327662e-01
## AccuracyPValue McnemarPValue
## 6.262798e-15 8.893031e-20
## model_id model_method
## 1 Conditional.X.rpart rpart
## feats
## 1 WordCount, PubDate.hour, PubDate.month, PubDate.date, PubDate.second, PubDate.minute, PubDate.wkday, S.year, A.year, S.compani, A.compani, S.time, A.time, S.will, A.will, S.york, A.york, S.new, A.new, H.new, S.week, A.week, SectionName.fctr, NewsDesk.fctr, SubsectionName.fctr, H.num.chars, H.num.chars.log, A.num.chars, S.num.chars, H.num.words, H.num.words.unq, H.num.words.log, S.num.chars.log, A.num.chars.log, A.num.words, S.num.words, H.num.words.unq.log, A.num.words.unq, S.num.words.unq, A.num.words.log, S.num.words.log, A.num.words.unq.log, S.num.words.unq.log
## max.nTuningRuns min.elapsedtime.everything min.elapsedtime.final
## 1 3 5.22 0.912
## max.auc.fit opt.prob.threshold.fit max.f.score.fit max.Accuracy.fit
## 1 0.830078 0.7 0.5738832 0.8893849
## max.AccuracyLower.fit max.AccuracyUpper.fit max.Kappa.fit max.auc.OOB
## 1 0.8795973 0.8982154 0.5222593 0.822503
## opt.prob.threshold.OOB max.f.score.OOB max.Accuracy.OOB
## 1 0.7 0.6014493 0.8930481
## max.AccuracyLower.OOB max.AccuracyUpper.OOB max.Kappa.OOB
## 1 0.878885 0.9060795 0.543976
## max.AccuracySD.fit max.KappaSD.fit
## 1 0.00245876 0.005834195
## [1] "fitting model: Conditional.X.cp.0.rpart"
## [1] " indep_vars: WordCount, PubDate.hour, PubDate.month, PubDate.date, PubDate.second, PubDate.minute, PubDate.wkday, S.year, A.year, S.compani, A.compani, S.time, A.time, S.will, A.will, S.york, A.york, S.new, A.new, H.new, S.week, A.week, SectionName.fctr, NewsDesk.fctr, SubsectionName.fctr, H.num.chars, H.num.chars.log, A.num.chars, S.num.chars, H.num.words, H.num.words.unq, H.num.words.log, S.num.chars.log, A.num.chars.log, A.num.words, S.num.words, H.num.words.unq.log, A.num.words.unq, S.num.words.unq, A.num.words.log, S.num.words.log, A.num.words.unq.log, S.num.words.unq.log"
## Fitting cp = 0 on full training set
## Call:
## rpart(formula = .outcome ~ ., control = list(minsplit = 20, minbucket = 7,
## cp = 0, maxcompete = 4, maxsurrogate = 5, usesurrogate = 2,
## surrogatestyle = 0, maxdepth = 30, xval = 0))
## n= 4475
##
## CP nsplit rel error
## 1 0.2590120160 0 1.0000000
## 2 0.0393858478 1 0.7409880
## 3 0.0280373832 3 0.6622163
## 4 0.0160213618 4 0.6341789
## 5 0.0140186916 6 0.6021362
## 6 0.0120160214 8 0.5740988
## 7 0.0093457944 10 0.5500668
## 8 0.0080106809 11 0.5407210
## 9 0.0045393858 12 0.5327103
## 10 0.0040053405 17 0.5100134
## 11 0.0033377837 20 0.4979973
## 12 0.0016021362 27 0.4739653
## 13 0.0013351135 49 0.4178905
## 14 0.0004450378 61 0.4018692
## 15 0.0000000000 67 0.3991989
##
## Variable importance
## NewsDesk.fctrOpEd SectionName.fctrOpinion
## 16 14
## WordCount A.num.chars
## 8 6
## S.num.chars S.num.chars.log
## 6 5
## A.num.chars.log PubDate.hour
## 5 5
## SectionName.fctrHealth NewsDesk.fctrScience
## 4 4
## H.num.chars SectionName.fctrU.S.
## 2 2
## PubDate.minute NewsDesk.fctrStyles
## 2 2
## H.num.words.unq H.num.words
## 2 2
## PubDate.wkday SectionName.fctr
## 2 2
## H.num.words.unq.log H.num.words.log
## 1 1
## H.num.chars.log A.num.words
## 1 1
## S.num.words PubDate.month
## 1 1
## PubDate.second A.num.words.unq
## 1 1
## S.num.words.unq PubDate.date
## 1 1
## SubsectionName.fctrThe Public Editor SubsectionName.fctrRoom For Debate
## 1 1
##
## Node number 1: 4475 observations, complexity param=0.259012
## predicted class=N expected loss=0.1673743 P(node) =1
## class counts: 3726 749
## probabilities: 0.833 0.167
## left son=2 (4131 obs) right son=3 (344 obs)
## Primary splits:
## NewsDesk.fctrOpEd < 0.5 to the left, improve=281.52360, (0 missing)
## SectionName.fctrOpinion < 0.5 to the left, improve=244.96950, (0 missing)
## WordCount < 683.5 to the left, improve=113.02630, (0 missing)
## A.num.chars < 44.5 to the right, improve= 93.81257, (0 missing)
## S.num.chars < 44.5 to the right, improve= 93.81257, (0 missing)
## Surrogate splits:
## SectionName.fctrOpinion < 0.5 to the left, agree=0.986, adj=0.817, (0 split)
## A.num.chars < 40.5 to the right, agree=0.933, adj=0.128, (0 split)
## S.num.chars < 40.5 to the right, agree=0.933, adj=0.128, (0 split)
## S.num.chars.log < 3.725621 to the right, agree=0.933, adj=0.128, (0 split)
## A.num.chars.log < 3.725621 to the right, agree=0.933, adj=0.128, (0 split)
##
## Node number 2: 4131 observations, complexity param=0.03938585
## predicted class=N expected loss=0.1161946 P(node) =0.9231285
## class counts: 3651 480
## probabilities: 0.884 0.116
## left son=4 (3066 obs) right son=5 (1065 obs)
## Primary splits:
## WordCount < 683.5 to the left, improve=105.55990, (0 missing)
## NewsDesk.fctrScience < 0.5 to the left, improve= 70.50963, (0 missing)
## SectionName.fctrHealth < 0.5 to the left, improve= 69.73481, (0 missing)
## PubDate.hour < 21.5 to the left, improve= 53.45620, (0 missing)
## NewsDesk.fctrStyles < 0.5 to the left, improve= 27.64076, (0 missing)
## Surrogate splits:
## SectionName.fctrOpinion < 0.5 to the left, agree=0.755, adj=0.048, (0 split)
## SubsectionName.fctrRoom For Debate < 0.5 to the left, agree=0.753, adj=0.041, (0 split)
## SectionName.fctrHealth < 0.5 to the left, agree=0.745, adj=0.012, (0 split)
## NewsDesk.fctrScience < 0.5 to the left, agree=0.745, adj=0.011, (0 split)
## SubsectionName.fctrThe Public Editor < 0.5 to the left, agree=0.744, adj=0.008, (0 split)
##
## Node number 3: 344 observations, complexity param=0.008010681
## predicted class=Y expected loss=0.2180233 P(node) =0.07687151
## class counts: 75 269
## probabilities: 0.218 0.782
## left son=6 (8 obs) right son=7 (336 obs)
## Primary splits:
## A.num.chars < 169.5 to the right, improve=7.070321, (0 missing)
## S.num.chars < 169.5 to the right, improve=7.070321, (0 missing)
## S.num.chars.log < 5.138628 to the right, improve=7.070321, (0 missing)
## A.num.chars.log < 5.138628 to the right, improve=7.070321, (0 missing)
## A.num.words < 15.5 to the right, improve=5.005347, (0 missing)
## Surrogate splits:
## S.num.chars < 169.5 to the right, agree=1.000, adj=1.000, (0 split)
## S.num.chars.log < 5.138628 to the right, agree=1.000, adj=1.000, (0 split)
## A.num.chars.log < 5.138628 to the right, agree=1.000, adj=1.000, (0 split)
## A.num.words < 17 to the right, agree=0.991, adj=0.625, (0 split)
## S.num.words < 17 to the right, agree=0.991, adj=0.625, (0 split)
##
## Node number 4: 3066 observations, complexity param=0.01602136
## predicted class=N expected loss=0.04957599 P(node) =0.6851397
## class counts: 2914 152
## probabilities: 0.950 0.050
## left son=8 (2987 obs) right son=9 (79 obs)
## Primary splits:
## PubDate.hour < 21.5 to the left, improve=23.51779, (0 missing)
## NewsDesk.fctrStyles < 0.5 to the left, improve=11.23429, (0 missing)
## A.num.chars < 44.5 to the right, improve=10.42832, (0 missing)
## S.num.chars < 44.5 to the right, improve=10.42832, (0 missing)
## S.num.chars.log < 3.817652 to the right, improve=10.42832, (0 missing)
##
## Node number 5: 1065 observations, complexity param=0.03938585
## predicted class=N expected loss=0.3079812 P(node) =0.2379888
## class counts: 737 328
## probabilities: 0.692 0.308
## left son=10 (994 obs) right son=11 (71 obs)
## Primary splits:
## SectionName.fctrHealth < 0.5 to the left, improve=56.15144, (0 missing)
## NewsDesk.fctrScience < 0.5 to the left, improve=55.08535, (0 missing)
## PubDate.hour < 7.5 to the left, improve=18.57852, (0 missing)
## NewsDesk.fctrForeign < 0.5 to the right, improve=11.91201, (0 missing)
## SectionName.fctrWorld < 0.5 to the right, improve=11.48588, (0 missing)
## Surrogate splits:
## NewsDesk.fctrScience < 0.5 to the left, agree=0.999, adj=0.986, (0 split)
## PubDate.hour < 0.5 to the right, agree=0.939, adj=0.085, (0 split)
##
## Node number 6: 8 observations
## predicted class=N expected loss=0.125 P(node) =0.001787709
## class counts: 7 1
## probabilities: 0.875 0.125
##
## Node number 7: 336 observations, complexity param=0.00400534
## predicted class=Y expected loss=0.202381 P(node) =0.0750838
## class counts: 68 268
## probabilities: 0.202 0.798
## left son=14 (34 obs) right son=15 (302 obs)
## Primary splits:
## WordCount < 107 to the left, improve=3.316861, (0 missing)
## PubDate.wkday < 3.5 to the right, improve=2.518525, (0 missing)
## PubDate.hour < 10.5 to the right, improve=2.041174, (0 missing)
## H.num.words.unq < 5.5 to the right, improve=1.827719, (0 missing)
## H.num.words.unq.log < 1.868835 to the right, improve=1.827719, (0 missing)
## Surrogate splits:
## H.num.chars < 9 to the left, agree=0.902, adj=0.029, (0 split)
## H.num.chars.log < 2.29756 to the left, agree=0.902, adj=0.029, (0 split)
##
## Node number 8: 2987 observations, complexity param=0.004539386
## predicted class=N expected loss=0.03950452 P(node) =0.667486
## class counts: 2869 118
## probabilities: 0.960 0.040
## left son=16 (2864 obs) right son=17 (123 obs)
## Primary splits:
## NewsDesk.fctrStyles < 0.5 to the left, improve=12.492140, (0 missing)
## NewsDesk.fctrScience < 0.5 to the left, improve= 7.607968, (0 missing)
## SectionName.fctrHealth < 0.5 to the left, improve= 6.608651, (0 missing)
## WordCount < 426.5 to the left, improve= 4.320240, (0 missing)
## SectionName.fctrU.S. < 0.5 to the left, improve= 4.093264, (0 missing)
## Surrogate splits:
## S.week < 1.5 to the left, agree=0.961, adj=0.057, (0 split)
## A.week < 1.5 to the left, agree=0.961, adj=0.057, (0 split)
## SectionName.fctrStyle < 0.5 to the left, agree=0.959, adj=0.016, (0 split)
## SubsectionName.fctrFashion & Style < 0.5 to the left, agree=0.959, adj=0.016, (0 split)
## A.num.chars < 379 to the left, agree=0.959, adj=0.008, (0 split)
##
## Node number 9: 79 observations, complexity param=0.01602136
## predicted class=N expected loss=0.4303797 P(node) =0.01765363
## class counts: 45 34
## probabilities: 0.570 0.430
## left son=18 (45 obs) right son=19 (34 obs)
## Primary splits:
## PubDate.minute < 0.5 to the right, improve=21.31588, (0 missing)
## A.num.chars < 62.5 to the right, improve=17.07763, (0 missing)
## S.num.chars < 62.5 to the right, improve=17.07763, (0 missing)
## S.num.chars.log < 4.151009 to the right, improve=17.07763, (0 missing)
## A.num.chars.log < 4.151009 to the right, improve=17.07763, (0 missing)
## Surrogate splits:
## H.num.words < 4.5 to the right, agree=0.861, adj=0.676, (0 split)
## H.num.words.unq < 4.5 to the right, agree=0.861, adj=0.676, (0 split)
## H.num.words.log < 1.700599 to the right, agree=0.861, adj=0.676, (0 split)
## H.num.words.unq.log < 1.700599 to the right, agree=0.861, adj=0.676, (0 split)
## H.num.chars < 34.5 to the right, agree=0.848, adj=0.647, (0 split)
##
## Node number 10: 994 observations, complexity param=0.02803738
## predicted class=N expected loss=0.2645875 P(node) =0.2221229
## class counts: 731 263
## probabilities: 0.735 0.265
## left son=20 (929 obs) right son=21 (65 obs)
## Primary splits:
## PubDate.hour < 21.5 to the left, improve=21.91733, (0 missing)
## A.num.chars < 78.5 to the right, improve=13.94732, (0 missing)
## S.num.chars < 78.5 to the right, improve=13.94732, (0 missing)
## S.num.chars.log < 4.375737 to the right, improve=13.94732, (0 missing)
## A.num.chars.log < 4.375737 to the right, improve=13.94732, (0 missing)
## Surrogate splits:
## SectionName.fctrArts < 0.5 to the left, agree=0.943, adj=0.123, (0 split)
## NewsDesk.fctrCulture < 0.5 to the left, agree=0.943, adj=0.123, (0 split)
## A.num.chars < 44 to the right, agree=0.943, adj=0.123, (0 split)
## S.num.chars < 44 to the right, agree=0.943, adj=0.123, (0 split)
## S.num.chars.log < 3.806416 to the right, agree=0.943, adj=0.123, (0 split)
##
## Node number 11: 71 observations
## predicted class=Y expected loss=0.08450704 P(node) =0.01586592
## class counts: 6 65
## probabilities: 0.085 0.915
##
## Node number 14: 34 observations, complexity param=0.00400534
## predicted class=Y expected loss=0.4117647 P(node) =0.007597765
## class counts: 14 20
## probabilities: 0.412 0.588
## left son=28 (23 obs) right son=29 (11 obs)
## Primary splits:
## PubDate.wkday < 2.5 to the right, improve=3.348059, (0 missing)
## PubDate.hour < 17.5 to the right, improve=3.313079, (0 missing)
## A.num.chars < 25.5 to the right, improve=2.228164, (0 missing)
## S.num.chars < 25.5 to the right, improve=2.228164, (0 missing)
## S.num.chars.log < 3.276967 to the right, improve=2.228164, (0 missing)
## Surrogate splits:
## PubDate.second < 12 to the right, agree=0.765, adj=0.273, (0 split)
## A.num.chars < 19.5 to the right, agree=0.765, adj=0.273, (0 split)
## S.num.chars < 19.5 to the right, agree=0.765, adj=0.273, (0 split)
## S.num.chars.log < 3.020127 to the right, agree=0.765, adj=0.273, (0 split)
## A.num.chars.log < 3.020127 to the right, agree=0.765, adj=0.273, (0 split)
##
## Node number 15: 302 observations, complexity param=0.001602136
## predicted class=Y expected loss=0.1788079 P(node) =0.06748603
## class counts: 54 248
## probabilities: 0.179 0.821
## left son=30 (254 obs) right son=31 (48 obs)
## Primary splits:
## A.num.chars < 38.5 to the right, improve=2.146747, (0 missing)
## S.num.chars < 38.5 to the right, improve=2.146747, (0 missing)
## S.num.chars.log < 3.676221 to the right, improve=2.146747, (0 missing)
## A.num.chars.log < 3.676221 to the right, improve=2.146747, (0 missing)
## H.num.chars < 21.5 to the right, improve=1.604170, (0 missing)
## Surrogate splits:
## S.num.chars < 38.5 to the right, agree=1.000, adj=1.000, (0 split)
## S.num.chars.log < 3.676221 to the right, agree=1.000, adj=1.000, (0 split)
## A.num.chars.log < 3.676221 to the right, agree=1.000, adj=1.000, (0 split)
## A.num.words.unq < 3.5 to the right, agree=0.937, adj=0.604, (0 split)
## S.num.words.unq < 3.5 to the right, agree=0.937, adj=0.604, (0 split)
##
## Node number 16: 2864 observations, complexity param=0.003337784
## predicted class=N expected loss=0.03002793 P(node) =0.64
## class counts: 2778 86
## probabilities: 0.970 0.030
## left son=32 (2806 obs) right son=33 (58 obs)
## Primary splits:
## NewsDesk.fctrScience < 0.5 to the left, improve=8.194155, (0 missing)
## SectionName.fctrHealth < 0.5 to the left, improve=7.308923, (0 missing)
## WordCount < 396.5 to the left, improve=2.632233, (0 missing)
## PubDate.wkday < 0.5 to the right, improve=2.360948, (0 missing)
## A.num.chars < 44.5 to the right, improve=2.305181, (0 missing)
## Surrogate splits:
## SectionName.fctrHealth < 0.5 to the left, agree=1, adj=0.983, (0 split)
##
## Node number 17: 123 observations, complexity param=0.004539386
## predicted class=N expected loss=0.2601626 P(node) =0.02748603
## class counts: 91 32
## probabilities: 0.740 0.260
## left son=34 (61 obs) right son=35 (62 obs)
## Primary splits:
## SectionName.fctrU.S. < 0.5 to the left, improve=16.381850, (0 missing)
## SectionName.fctr < 0.5 to the right, improve=14.857290, (0 missing)
## A.num.chars < 127.5 to the left, improve= 5.191844, (0 missing)
## S.num.chars < 127.5 to the left, improve= 5.191844, (0 missing)
## S.num.chars.log < 4.855921 to the left, improve= 5.191844, (0 missing)
## Surrogate splits:
## SectionName.fctr < 0.5 to the right, agree=0.976, adj=0.951, (0 split)
## PubDate.hour < 9.5 to the left, agree=0.732, adj=0.459, (0 split)
## PubDate.month < 9.5 to the left, agree=0.699, adj=0.393, (0 split)
## H.num.chars < 45.5 to the left, agree=0.699, adj=0.393, (0 split)
## H.num.chars.log < 3.839394 to the left, agree=0.699, adj=0.393, (0 split)
##
## Node number 18: 45 observations
## predicted class=N expected loss=0.1111111 P(node) =0.01005587
## class counts: 40 5
## probabilities: 0.889 0.111
##
## Node number 19: 34 observations, complexity param=0.001335113
## predicted class=Y expected loss=0.1470588 P(node) =0.007597765
## class counts: 5 29
## probabilities: 0.147 0.853
## left son=38 (7 obs) right son=39 (27 obs)
## Primary splits:
## WordCount < 327.5 to the left, improve=3.174914, (0 missing)
## H.num.chars < 29.5 to the right, improve=3.174914, (0 missing)
## H.num.chars.log < 3.417592 to the right, improve=3.174914, (0 missing)
## A.num.chars < 82.5 to the right, improve=3.174914, (0 missing)
## S.num.chars < 82.5 to the right, improve=3.174914, (0 missing)
## Surrogate splits:
## PubDate.wkday < 0.5 to the left, agree=0.882, adj=0.429, (0 split)
## SectionName.fctrN.Y. / Region < 0.5 to the right, agree=0.882, adj=0.429, (0 split)
## NewsDesk.fctrMetro < 0.5 to the right, agree=0.882, adj=0.429, (0 split)
## A.num.chars < 127 to the right, agree=0.853, adj=0.286, (0 split)
## S.num.chars < 127 to the right, agree=0.853, adj=0.286, (0 split)
##
## Node number 20: 929 observations, complexity param=0.01401869
## predicted class=N expected loss=0.2368138 P(node) =0.2075978
## class counts: 709 220
## probabilities: 0.763 0.237
## left son=40 (851 obs) right son=41 (78 obs)
## Primary splits:
## NewsDesk.fctrStyles < 0.5 to the left, improve=16.840870, (0 missing)
## SectionName.fctrU.S. < 0.5 to the left, improve=13.129340, (0 missing)
## PubDate.hour < 7.5 to the left, improve=12.747800, (0 missing)
## PubDate.wkday < 5.5 to the left, improve= 8.649968, (0 missing)
## SubsectionName.fctrThe Public Editor < 0.5 to the left, improve= 7.524853, (0 missing)
## Surrogate splits:
## SectionName.fctrU.S. < 0.5 to the left, agree=0.963, adj=0.564, (0 split)
##
## Node number 21: 65 observations, complexity param=0.01201602
## predicted class=Y expected loss=0.3384615 P(node) =0.01452514
## class counts: 22 43
## probabilities: 0.338 0.662
## left son=42 (25 obs) right son=43 (40 obs)
## Primary splits:
## A.num.chars < 92 to the right, improve=9.477692, (0 missing)
## S.num.chars < 92 to the right, improve=9.477692, (0 missing)
## S.num.chars.log < 4.532368 to the right, improve=9.477692, (0 missing)
## A.num.chars.log < 4.532368 to the right, improve=9.477692, (0 missing)
## A.num.words.unq < 9.5 to the right, improve=7.387692, (0 missing)
## Surrogate splits:
## S.num.chars < 92 to the right, agree=1.000, adj=1.00, (0 split)
## S.num.chars.log < 4.532368 to the right, agree=1.000, adj=1.00, (0 split)
## A.num.chars.log < 4.532368 to the right, agree=1.000, adj=1.00, (0 split)
## A.num.words.unq < 9.5 to the right, agree=0.969, adj=0.92, (0 split)
## S.num.words.unq < 9.5 to the right, agree=0.969, adj=0.92, (0 split)
##
## Node number 28: 23 observations, complexity param=0.00400534
## predicted class=N expected loss=0.4347826 P(node) =0.005139665
## class counts: 13 10
## probabilities: 0.565 0.435
## left son=56 (11 obs) right son=57 (12 obs)
## Primary splits:
## PubDate.hour < 15.5 to the right, improve=4.986166, (0 missing)
## WordCount < 53 to the right, improve=1.715062, (0 missing)
## PubDate.second < 31.5 to the left, improve=1.715062, (0 missing)
## A.num.chars < 27.5 to the right, improve=1.590062, (0 missing)
## S.num.chars < 27.5 to the right, improve=1.590062, (0 missing)
## Surrogate splits:
## A.num.chars < 25.5 to the right, agree=0.783, adj=0.545, (0 split)
## S.num.chars < 25.5 to the right, agree=0.783, adj=0.545, (0 split)
## S.num.chars.log < 3.276967 to the right, agree=0.783, adj=0.545, (0 split)
## A.num.chars.log < 3.276967 to the right, agree=0.783, adj=0.545, (0 split)
## WordCount < 53 to the right, agree=0.652, adj=0.273, (0 split)
##
## Node number 29: 11 observations
## predicted class=Y expected loss=0.09090909 P(node) =0.002458101
## class counts: 1 10
## probabilities: 0.091 0.909
##
## Node number 30: 254 observations, complexity param=0.001602136
## predicted class=Y expected loss=0.2047244 P(node) =0.05675978
## class counts: 52 202
## probabilities: 0.205 0.795
## left son=60 (105 obs) right son=61 (149 obs)
## Primary splits:
## PubDate.wkday < 3.5 to the right, improve=1.373538, (0 missing)
## PubDate.minute < 32.5 to the left, improve=1.202827, (0 missing)
## PubDate.second < 51.5 to the left, improve=1.149888, (0 missing)
## WordCount < 352.5 to the left, improve=1.099088, (0 missing)
## H.num.chars < 21.5 to the right, improve=1.031389, (0 missing)
## Surrogate splits:
## WordCount < 303.5 to the left, agree=0.618, adj=0.076, (0 split)
## H.num.chars < 48.5 to the right, agree=0.610, adj=0.057, (0 split)
## H.num.chars.log < 3.901922 to the right, agree=0.610, adj=0.057, (0 split)
## A.num.chars < 138.5 to the right, agree=0.606, adj=0.048, (0 split)
## S.num.chars < 138.5 to the right, agree=0.606, adj=0.048, (0 split)
##
## Node number 31: 48 observations
## predicted class=Y expected loss=0.04166667 P(node) =0.01072626
## class counts: 2 46
## probabilities: 0.042 0.958
##
## Node number 32: 2806 observations, complexity param=0.003337784
## predicted class=N expected loss=0.02459016 P(node) =0.6270391
## class counts: 2737 69
## probabilities: 0.975 0.025
## left son=64 (2102 obs) right son=65 (704 obs)
## Primary splits:
## WordCount < 396.5 to the left, improve=2.907448, (0 missing)
## PubDate.wkday < 0.5 to the right, improve=2.650636, (0 missing)
## A.num.chars < 44.5 to the right, improve=2.440507, (0 missing)
## S.num.chars < 44.5 to the right, improve=2.440507, (0 missing)
## S.num.chars.log < 3.817652 to the right, improve=2.440507, (0 missing)
## Surrogate splits:
## SectionName.fctrBusiness Day < 0.5 to the left, agree=0.762, adj=0.053, (0 split)
## SectionName.fctrWorld < 0.5 to the left, agree=0.762, adj=0.053, (0 split)
## SubsectionName.fctrAsia Pacific < 0.5 to the left, agree=0.762, adj=0.053, (0 split)
## SubsectionName.fctrSmall Business < 0.5 to the left, agree=0.762, adj=0.050, (0 split)
## H.num.words.unq < 1.5 to the right, agree=0.755, adj=0.023, (0 split)
##
## Node number 33: 58 observations, complexity param=0.003337784
## predicted class=N expected loss=0.2931034 P(node) =0.01296089
## class counts: 41 17
## probabilities: 0.707 0.293
## left son=66 (48 obs) right son=67 (10 obs)
## Primary splits:
## PubDate.hour < 9.5 to the right, improve=4.0011490, (0 missing)
## PubDate.wkday < 3.5 to the left, improve=3.0344830, (0 missing)
## H.num.chars < 46.5 to the right, improve=1.8137040, (0 missing)
## H.num.chars.log < 3.860674 to the right, improve=1.8137040, (0 missing)
## A.num.chars < 129 to the left, improve=0.9536747, (0 missing)
## Surrogate splits:
## PubDate.second < 2.5 to the right, agree=0.845, adj=0.1, (0 split)
##
## Node number 34: 61 observations
## predicted class=N expected loss=0 P(node) =0.01363128
## class counts: 61 0
## probabilities: 1.000 0.000
##
## Node number 35: 62 observations, complexity param=0.004539386
## predicted class=Y expected loss=0.483871 P(node) =0.01385475
## class counts: 30 32
## probabilities: 0.484 0.516
## left son=70 (53 obs) right son=71 (9 obs)
## Primary splits:
## PubDate.second < 8.5 to the right, improve=2.925813, (0 missing)
## H.num.chars < 55.5 to the left, improve=2.669088, (0 missing)
## H.num.chars.log < 4.034201 to the left, improve=2.669088, (0 missing)
## PubDate.minute < 51.5 to the left, improve=2.365890, (0 missing)
## H.num.words < 5.5 to the left, improve=2.249904, (0 missing)
## Surrogate splits:
## PubDate.minute < 57.5 to the left, agree=0.871, adj=0.111, (0 split)
##
## Node number 38: 7 observations
## predicted class=N expected loss=0.4285714 P(node) =0.001564246
## class counts: 4 3
## probabilities: 0.571 0.429
##
## Node number 39: 27 observations
## predicted class=Y expected loss=0.03703704 P(node) =0.00603352
## class counts: 1 26
## probabilities: 0.037 0.963
##
## Node number 40: 851 observations, complexity param=0.01201602
## predicted class=N expected loss=0.2079906 P(node) =0.1901676
## class counts: 674 177
## probabilities: 0.792 0.208
## left son=80 (834 obs) right son=81 (17 obs)
## Primary splits:
## PubDate.wkday < 5.5 to the left, improve=10.752480, (0 missing)
## SubsectionName.fctrThe Public Editor < 0.5 to the left, improve= 8.298601, (0 missing)
## PubDate.hour < 7.5 to the left, improve= 7.895594, (0 missing)
## H.num.chars < 16.5 to the right, improve= 6.622781, (0 missing)
## H.num.chars.log < 2.861793 to the right, improve= 6.622781, (0 missing)
## Surrogate splits:
## H.num.chars < 8.5 to the right, agree=0.982, adj=0.118, (0 split)
## H.num.chars.log < 2.249905 to the right, agree=0.982, adj=0.118, (0 split)
##
## Node number 41: 78 observations, complexity param=0.01401869
## predicted class=Y expected loss=0.4487179 P(node) =0.01743017
## class counts: 35 43
## probabilities: 0.449 0.551
## left son=82 (13 obs) right son=83 (65 obs)
## Primary splits:
## SectionName.fctr < 0.5 to the right, improve=9.482051, (0 missing)
## SectionName.fctrU.S. < 0.5 to the left, improve=9.482051, (0 missing)
## A.num.chars < 174.5 to the left, improve=5.128205, (0 missing)
## S.num.chars < 174.5 to the left, improve=5.128205, (0 missing)
## S.num.chars.log < 5.167635 to the left, improve=5.128205, (0 missing)
## Surrogate splits:
## SectionName.fctrU.S. < 0.5 to the left, agree=1.000, adj=1.000, (0 split)
## S.week < 0.5 to the right, agree=0.859, adj=0.154, (0 split)
## A.week < 0.5 to the right, agree=0.859, adj=0.154, (0 split)
## WordCount < 715.5 to the left, agree=0.846, adj=0.077, (0 split)
## PubDate.minute < 57.5 to the right, agree=0.846, adj=0.077, (0 split)
##
## Node number 42: 25 observations, complexity param=0.001335113
## predicted class=N expected loss=0.32 P(node) =0.005586592
## class counts: 17 8
## probabilities: 0.680 0.320
## left son=84 (18 obs) right son=85 (7 obs)
## Primary splits:
## PubDate.date < 11 to the right, improve=1.229206, (0 missing)
## H.num.words < 4.5 to the right, improve=1.229206, (0 missing)
## H.num.words.log < 1.700599 to the right, improve=1.229206, (0 missing)
## A.num.words < 16.5 to the left, improve=1.229206, (0 missing)
## S.num.words < 16.5 to the left, improve=1.229206, (0 missing)
## Surrogate splits:
## A.num.chars < 178.5 to the left, agree=0.8, adj=0.286, (0 split)
## S.num.chars < 178.5 to the left, agree=0.8, adj=0.286, (0 split)
## S.num.chars.log < 5.188461 to the left, agree=0.8, adj=0.286, (0 split)
## A.num.chars.log < 5.188461 to the left, agree=0.8, adj=0.286, (0 split)
##
## Node number 43: 40 observations
## predicted class=Y expected loss=0.125 P(node) =0.008938547
## class counts: 5 35
## probabilities: 0.125 0.875
##
## Node number 56: 11 observations
## predicted class=N expected loss=0.09090909 P(node) =0.002458101
## class counts: 10 1
## probabilities: 0.909 0.091
##
## Node number 57: 12 observations
## predicted class=Y expected loss=0.25 P(node) =0.002681564
## class counts: 3 9
## probabilities: 0.250 0.750
##
## Node number 60: 105 observations, complexity param=0.001602136
## predicted class=Y expected loss=0.2666667 P(node) =0.02346369
## class counts: 28 77
## probabilities: 0.267 0.733
## left son=120 (48 obs) right son=121 (57 obs)
## Primary splits:
## WordCount < 483 to the left, improve=2.075439, (0 missing)
## PubDate.wkday < 5.5 to the left, improve=1.747518, (0 missing)
## PubDate.minute < 12.5 to the right, improve=1.542857, (0 missing)
## PubDate.second < 47.5 to the left, improve=1.093178, (0 missing)
## H.num.words < 2.5 to the right, improve=1.068339, (0 missing)
## Surrogate splits:
## PubDate.minute < 53.5 to the right, agree=0.629, adj=0.188, (0 split)
## PubDate.date < 14.5 to the left, agree=0.610, adj=0.146, (0 split)
## PubDate.second < 54.5 to the right, agree=0.590, adj=0.104, (0 split)
## H.num.chars < 54.5 to the right, agree=0.581, adj=0.083, (0 split)
## H.num.chars.log < 4.016342 to the right, agree=0.581, adj=0.083, (0 split)
##
## Node number 61: 149 observations
## predicted class=Y expected loss=0.1610738 P(node) =0.03329609
## class counts: 24 125
## probabilities: 0.161 0.839
##
## Node number 64: 2102 observations
## predicted class=N expected loss=0.0114177 P(node) =0.4697207
## class counts: 2078 24
## probabilities: 0.989 0.011
##
## Node number 65: 704 observations, complexity param=0.003337784
## predicted class=N expected loss=0.06392045 P(node) =0.1573184
## class counts: 659 45
## probabilities: 0.936 0.064
## left son=130 (694 obs) right son=131 (10 obs)
## Primary splits:
## A.num.chars < 56.5 to the right, improve=10.992400, (0 missing)
## S.num.chars < 56.5 to the right, improve=10.992400, (0 missing)
## S.num.chars.log < 4.051747 to the right, improve=10.992400, (0 missing)
## A.num.chars.log < 4.051747 to the right, improve=10.992400, (0 missing)
## PubDate.wkday < 0.5 to the right, improve= 6.557416, (0 missing)
## Surrogate splits:
## S.num.chars < 56.5 to the right, agree=1.00, adj=1.0, (0 split)
## S.num.chars.log < 4.051747 to the right, agree=1.00, adj=1.0, (0 split)
## A.num.chars.log < 4.051747 to the right, agree=1.00, adj=1.0, (0 split)
## A.num.words < 4.5 to the right, agree=0.99, adj=0.3, (0 split)
## S.num.words < 4.5 to the right, agree=0.99, adj=0.3, (0 split)
##
## Node number 66: 48 observations
## predicted class=N expected loss=0.2083333 P(node) =0.01072626
## class counts: 38 10
## probabilities: 0.792 0.208
##
## Node number 67: 10 observations
## predicted class=Y expected loss=0.3 P(node) =0.002234637
## class counts: 3 7
## probabilities: 0.300 0.700
##
## Node number 70: 53 observations, complexity param=0.004539386
## predicted class=N expected loss=0.4528302 P(node) =0.01184358
## class counts: 29 24
## probabilities: 0.547 0.453
## left son=140 (16 obs) right son=141 (37 obs)
## Primary splits:
## S.week < 0.5 to the right, improve=3.226989, (0 missing)
## A.week < 0.5 to the right, improve=3.226989, (0 missing)
## H.num.words < 5.5 to the left, improve=2.533946, (0 missing)
## H.num.words.unq < 5.5 to the left, improve=2.533946, (0 missing)
## H.num.words.log < 1.868835 to the left, improve=2.533946, (0 missing)
## Surrogate splits:
## A.week < 0.5 to the right, agree=1.000, adj=1.000, (0 split)
## WordCount < 52 to the left, agree=0.887, adj=0.625, (0 split)
## PubDate.wkday < 4.5 to the right, agree=0.849, adj=0.500, (0 split)
## H.num.chars < 48 to the left, agree=0.811, adj=0.375, (0 split)
## H.num.chars.log < 3.891612 to the left, agree=0.811, adj=0.375, (0 split)
##
## Node number 71: 9 observations
## predicted class=Y expected loss=0.1111111 P(node) =0.002011173
## class counts: 1 8
## probabilities: 0.111 0.889
##
## Node number 80: 834 observations, complexity param=0.009345794
## predicted class=N expected loss=0.1966427 P(node) =0.1863687
## class counts: 670 164
## probabilities: 0.803 0.197
## left son=160 (823 obs) right son=161 (11 obs)
## Primary splits:
## SubsectionName.fctrThe Public Editor < 0.5 to the left, improve=8.612433, (0 missing)
## PubDate.hour < 7.5 to the left, improve=6.615481, (0 missing)
## NewsDesk.fctrForeign < 0.5 to the right, improve=4.228906, (0 missing)
## SectionName.fctrWorld < 0.5 to the right, improve=4.141039, (0 missing)
## A.num.chars < 188.5 to the right, improve=3.859047, (0 missing)
##
## Node number 81: 17 observations
## predicted class=Y expected loss=0.2352941 P(node) =0.003798883
## class counts: 4 13
## probabilities: 0.235 0.765
##
## Node number 82: 13 observations
## predicted class=N expected loss=0 P(node) =0.002905028
## class counts: 13 0
## probabilities: 1.000 0.000
##
## Node number 83: 65 observations, complexity param=0.003337784
## predicted class=Y expected loss=0.3384615 P(node) =0.01452514
## class counts: 22 43
## probabilities: 0.338 0.662
## left son=166 (40 obs) right son=167 (25 obs)
## Primary splits:
## A.num.chars < 174.5 to the left, improve=2.587692, (0 missing)
## S.num.chars < 174.5 to the left, improve=2.587692, (0 missing)
## S.num.chars.log < 5.167635 to the left, improve=2.587692, (0 missing)
## A.num.chars.log < 5.167635 to the left, improve=2.587692, (0 missing)
## S.new < 0.5 to the right, improve=2.216067, (0 missing)
## Surrogate splits:
## S.num.chars < 174.5 to the left, agree=1.000, adj=1.00, (0 split)
## S.num.chars.log < 5.167635 to the left, agree=1.000, adj=1.00, (0 split)
## A.num.chars.log < 5.167635 to the left, agree=1.000, adj=1.00, (0 split)
## A.num.words < 15.5 to the left, agree=0.862, adj=0.64, (0 split)
## S.num.words < 15.5 to the left, agree=0.862, adj=0.64, (0 split)
##
## Node number 84: 18 observations
## predicted class=N expected loss=0.2222222 P(node) =0.004022346
## class counts: 14 4
## probabilities: 0.778 0.222
##
## Node number 85: 7 observations
## predicted class=Y expected loss=0.4285714 P(node) =0.001564246
## class counts: 3 4
## probabilities: 0.429 0.571
##
## Node number 120: 48 observations, complexity param=0.001602136
## predicted class=Y expected loss=0.375 P(node) =0.01072626
## class counts: 18 30
## probabilities: 0.375 0.625
## left son=240 (41 obs) right son=241 (7 obs)
## Primary splits:
## A.num.chars < 56 to the right, improve=2.304878, (0 missing)
## S.num.chars < 56 to the right, improve=2.304878, (0 missing)
## S.num.chars.log < 4.042897 to the right, improve=2.304878, (0 missing)
## A.num.chars.log < 4.042897 to the right, improve=2.304878, (0 missing)
## PubDate.date < 24.5 to the right, improve=1.886760, (0 missing)
## Surrogate splits:
## S.num.chars < 56 to the right, agree=1.000, adj=1.000, (0 split)
## S.num.chars.log < 4.042897 to the right, agree=1.000, adj=1.000, (0 split)
## A.num.chars.log < 4.042897 to the right, agree=1.000, adj=1.000, (0 split)
## A.num.words.unq < 5.5 to the right, agree=0.938, adj=0.571, (0 split)
## S.num.words.unq < 5.5 to the right, agree=0.938, adj=0.571, (0 split)
##
## Node number 121: 57 observations
## predicted class=Y expected loss=0.1754386 P(node) =0.01273743
## class counts: 10 47
## probabilities: 0.175 0.825
##
## Node number 130: 694 observations, complexity param=0.0004450378
## predicted class=N expected loss=0.05331412 P(node) =0.1550838
## class counts: 657 37
## probabilities: 0.947 0.053
## left son=260 (250 obs) right son=261 (444 obs)
## Primary splits:
## SectionName.fctrBusiness Day < 0.5 to the right, improve=1.9005930, (0 missing)
## SubsectionName.fctrDealbook < 0.5 to the right, improve=1.3867990, (0 missing)
## PubDate.hour < 18.5 to the left, improve=1.2015070, (0 missing)
## SectionName.fctrTechnology < 0.5 to the left, improve=1.1521010, (0 missing)
## PubDate.wkday < 0.5 to the right, improve=0.8343778, (0 missing)
## Surrogate splits:
## SubsectionName.fctrDealbook < 0.5 to the right, agree=0.938, adj=0.828, (0 split)
## SubsectionName.fctrSmall Business < 0.5 to the right, agree=0.702, adj=0.172, (0 split)
## S.compani < 0.5 to the right, agree=0.689, adj=0.136, (0 split)
## A.compani < 0.5 to the right, agree=0.689, adj=0.136, (0 split)
## WordCount < 411.5 to the left, agree=0.641, adj=0.004, (0 split)
##
## Node number 131: 10 observations
## predicted class=Y expected loss=0.2 P(node) =0.002234637
## class counts: 2 8
## probabilities: 0.200 0.800
##
## Node number 140: 16 observations
## predicted class=N expected loss=0.1875 P(node) =0.003575419
## class counts: 13 3
## probabilities: 0.812 0.188
##
## Node number 141: 37 observations, complexity param=0.004539386
## predicted class=Y expected loss=0.4324324 P(node) =0.008268156
## class counts: 16 21
## probabilities: 0.432 0.568
## left son=282 (9 obs) right son=283 (28 obs)
## Primary splits:
## PubDate.second < 21.5 to the left, improve=2.836765, (0 missing)
## WordCount < 310 to the left, improve=1.305019, (0 missing)
## PubDate.month < 10.5 to the right, improve=1.305019, (0 missing)
## A.num.chars < 199 to the right, improve=1.305019, (0 missing)
## S.num.chars < 199 to the right, improve=1.305019, (0 missing)
## Surrogate splits:
## WordCount < 190 to the left, agree=0.811, adj=0.222, (0 split)
## PubDate.minute < 33.5 to the right, agree=0.811, adj=0.222, (0 split)
## A.num.words < 24.5 to the right, agree=0.811, adj=0.222, (0 split)
## S.num.words < 23.5 to the right, agree=0.811, adj=0.222, (0 split)
## A.num.words.log < 3.238486 to the right, agree=0.811, adj=0.222, (0 split)
##
## Node number 160: 823 observations, complexity param=0.001602136
## predicted class=N expected loss=0.1883354 P(node) =0.1839106
## class counts: 668 155
## probabilities: 0.812 0.188
## left son=320 (267 obs) right son=321 (556 obs)
## Primary splits:
## PubDate.hour < 7.5 to the left, improve=5.506684, (0 missing)
## SectionName.fctr < 0.5 to the left, improve=4.475767, (0 missing)
## WordCount < 3668 to the left, improve=3.905955, (0 missing)
## NewsDesk.fctrForeign < 0.5 to the right, improve=3.824819, (0 missing)
## SectionName.fctrWorld < 0.5 to the right, improve=3.744282, (0 missing)
## Surrogate splits:
## SectionName.fctrN.Y. / Region < 0.5 to the right, agree=0.724, adj=0.150, (0 split)
## NewsDesk.fctrMetro < 0.5 to the right, agree=0.724, adj=0.150, (0 split)
## SectionName.fctrMultimedia < 0.5 to the right, agree=0.713, adj=0.116, (0 split)
## H.new < 0.5 to the right, agree=0.707, adj=0.097, (0 split)
## SubsectionName.fctrSmall Business < 0.5 to the right, agree=0.706, adj=0.094, (0 split)
##
## Node number 161: 11 observations
## predicted class=Y expected loss=0.1818182 P(node) =0.002458101
## class counts: 2 9
## probabilities: 0.182 0.818
##
## Node number 166: 40 observations, complexity param=0.003337784
## predicted class=Y expected loss=0.45 P(node) =0.008938547
## class counts: 18 22
## probabilities: 0.450 0.550
## left son=332 (17 obs) right son=333 (23 obs)
## Primary splits:
## PubDate.date < 17.5 to the right, improve=2.296164, (0 missing)
## A.num.words.unq < 10.5 to the right, improve=2.182445, (0 missing)
## S.num.words.unq < 10.5 to the right, improve=2.182445, (0 missing)
## A.num.words.unq.log < 2.441401 to the right, improve=2.182445, (0 missing)
## S.num.words.unq.log < 2.441401 to the right, improve=2.182445, (0 missing)
## Surrogate splits:
## A.num.chars < 141.5 to the right, agree=0.700, adj=0.294, (0 split)
## S.num.chars < 141.5 to the right, agree=0.700, adj=0.294, (0 split)
## S.num.chars.log < 4.959336 to the right, agree=0.700, adj=0.294, (0 split)
## A.num.chars.log < 4.959336 to the right, agree=0.700, adj=0.294, (0 split)
## PubDate.second < 29 to the left, agree=0.675, adj=0.235, (0 split)
##
## Node number 167: 25 observations
## predicted class=Y expected loss=0.16 P(node) =0.005586592
## class counts: 4 21
## probabilities: 0.160 0.840
##
## Node number 240: 41 observations, complexity param=0.001602136
## predicted class=Y expected loss=0.4390244 P(node) =0.009162011
## class counts: 18 23
## probabilities: 0.439 0.561
## left son=480 (10 obs) right son=481 (31 obs)
## Primary splits:
## A.num.chars < 70.5 to the left, improve=3.446735, (0 missing)
## S.num.chars < 70.5 to the left, improve=3.446735, (0 missing)
## S.num.chars.log < 4.268498 to the left, improve=3.446735, (0 missing)
## A.num.chars.log < 4.268498 to the left, improve=3.446735, (0 missing)
## H.num.words.unq < 4.5 to the left, improve=2.513749, (0 missing)
## Surrogate splits:
## S.num.chars < 70.5 to the left, agree=1.000, adj=1.0, (0 split)
## S.num.chars.log < 4.268498 to the left, agree=1.000, adj=1.0, (0 split)
## A.num.chars.log < 4.268498 to the left, agree=1.000, adj=1.0, (0 split)
## A.num.words < 6.5 to the left, agree=0.854, adj=0.4, (0 split)
## S.num.words < 6.5 to the left, agree=0.854, adj=0.4, (0 split)
##
## Node number 241: 7 observations
## predicted class=Y expected loss=0 P(node) =0.001564246
## class counts: 0 7
## probabilities: 0.000 1.000
##
## Node number 260: 250 observations
## predicted class=N expected loss=0.004 P(node) =0.05586592
## class counts: 249 1
## probabilities: 0.996 0.004
##
## Node number 261: 444 observations, complexity param=0.0004450378
## predicted class=N expected loss=0.08108108 P(node) =0.09921788
## class counts: 408 36
## probabilities: 0.919 0.081
## left son=522 (415 obs) right son=523 (29 obs)
## Primary splits:
## PubDate.hour < 18.5 to the left, improve=1.5944850, (0 missing)
## PubDate.date < 23.5 to the left, improve=1.1176040, (0 missing)
## NewsDesk.fctrTStyle < 0.5 to the right, improve=1.0193050, (0 missing)
## PubDate.wkday < 0.5 to the right, improve=0.8285489, (0 missing)
## NewsDesk.fctrForeign < 0.5 to the right, improve=0.8254012, (0 missing)
##
## Node number 282: 9 observations
## predicted class=N expected loss=0.2222222 P(node) =0.002011173
## class counts: 7 2
## probabilities: 0.778 0.222
##
## Node number 283: 28 observations
## predicted class=Y expected loss=0.3214286 P(node) =0.006256983
## class counts: 9 19
## probabilities: 0.321 0.679
##
## Node number 320: 267 observations, complexity param=0.001335113
## predicted class=N expected loss=0.1048689 P(node) =0.0596648
## class counts: 239 28
## probabilities: 0.895 0.105
## left son=640 (252 obs) right son=641 (15 obs)
## Primary splits:
## S.compani < 0.5 to the left, improve=1.659087, (0 missing)
## A.compani < 0.5 to the left, improve=1.659087, (0 missing)
## SectionName.fctr < 0.5 to the left, improve=1.400771, (0 missing)
## PubDate.hour < 2.5 to the right, improve=1.300517, (0 missing)
## PubDate.minute < 5.5 to the left, improve=1.050223, (0 missing)
## Surrogate splits:
## A.compani < 0.5 to the left, agree=1.000, adj=1.000, (0 split)
## PubDate.hour < 0.5 to the right, agree=0.948, adj=0.067, (0 split)
##
## Node number 321: 556 observations, complexity param=0.001602136
## predicted class=N expected loss=0.2284173 P(node) =0.1242458
## class counts: 429 127
## probabilities: 0.772 0.228
## left son=642 (94 obs) right son=643 (462 obs)
## Primary splits:
## A.num.chars < 188.5 to the right, improve=3.982475, (0 missing)
## S.num.chars < 188.5 to the right, improve=3.982475, (0 missing)
## S.num.chars.log < 5.244386 to the right, improve=3.982475, (0 missing)
## A.num.chars.log < 5.244386 to the right, improve=3.982475, (0 missing)
## A.num.words.unq < 17.5 to the right, improve=3.822076, (0 missing)
## Surrogate splits:
## S.num.chars < 188.5 to the right, agree=1.000, adj=1.000, (0 split)
## S.num.chars.log < 5.244386 to the right, agree=1.000, adj=1.000, (0 split)
## A.num.chars.log < 5.244386 to the right, agree=1.000, adj=1.000, (0 split)
## A.num.words < 19.5 to the right, agree=0.928, adj=0.574, (0 split)
## A.num.words.log < 3.020127 to the right, agree=0.928, adj=0.574, (0 split)
##
## Node number 332: 17 observations
## predicted class=N expected loss=0.3529412 P(node) =0.003798883
## class counts: 11 6
## probabilities: 0.647 0.353
##
## Node number 333: 23 observations, complexity param=0.003337784
## predicted class=Y expected loss=0.3043478 P(node) =0.005139665
## class counts: 7 16
## probabilities: 0.304 0.696
## left son=666 (7 obs) right son=667 (16 obs)
## Primary splits:
## PubDate.minute < 38 to the right, improve=3.381988, (0 missing)
## PubDate.second < 33.5 to the left, improve=3.092977, (0 missing)
## A.num.words.unq < 10.5 to the right, improve=2.272464, (0 missing)
## S.num.words.unq < 10.5 to the right, improve=2.272464, (0 missing)
## A.num.words.unq.log < 2.441401 to the right, improve=2.272464, (0 missing)
## Surrogate splits:
## PubDate.hour < 17.5 to the right, agree=0.783, adj=0.286, (0 split)
## H.num.words < 2.5 to the left, agree=0.783, adj=0.286, (0 split)
## H.num.words.unq < 2.5 to the left, agree=0.783, adj=0.286, (0 split)
## H.num.words.log < 1.242453 to the left, agree=0.783, adj=0.286, (0 split)
## H.num.words.unq.log < 1.242453 to the left, agree=0.783, adj=0.286, (0 split)
##
## Node number 480: 10 observations
## predicted class=N expected loss=0.2 P(node) =0.002234637
## class counts: 8 2
## probabilities: 0.800 0.200
##
## Node number 481: 31 observations, complexity param=0.001602136
## predicted class=Y expected loss=0.3225806 P(node) =0.006927374
## class counts: 10 21
## probabilities: 0.323 0.677
## left son=962 (12 obs) right son=963 (19 obs)
## Primary splits:
## H.num.chars < 39 to the left, improve=2.662422, (0 missing)
## H.num.chars.log < 3.688567 to the left, improve=2.662422, (0 missing)
## H.num.words < 4.5 to the left, improve=2.662422, (0 missing)
## H.num.words.unq < 4.5 to the left, improve=2.662422, (0 missing)
## H.num.words.log < 1.700599 to the left, improve=2.662422, (0 missing)
## Surrogate splits:
## H.num.chars.log < 3.688567 to the left, agree=1.000, adj=1.000, (0 split)
## H.num.words < 4.5 to the left, agree=0.935, adj=0.833, (0 split)
## H.num.words.unq < 4.5 to the left, agree=0.935, adj=0.833, (0 split)
## H.num.words.log < 1.700599 to the left, agree=0.935, adj=0.833, (0 split)
## H.num.words.unq.log < 1.700599 to the left, agree=0.935, adj=0.833, (0 split)
##
## Node number 522: 415 observations, complexity param=0.0004450378
## predicted class=N expected loss=0.06987952 P(node) =0.09273743
## class counts: 386 29
## probabilities: 0.930 0.070
## left son=1044 (316 obs) right son=1045 (99 obs)
## Primary splits:
## PubDate.date < 23.5 to the left, improve=1.3306350, (0 missing)
## NewsDesk.fctrTStyle < 0.5 to the right, improve=0.7390107, (0 missing)
## PubDate.second < 9.5 to the right, improve=0.7042069, (0 missing)
## SectionName.fctrTechnology < 0.5 to the left, improve=0.6535238, (0 missing)
## NewsDesk.fctrForeign < 0.5 to the right, improve=0.5933968, (0 missing)
## Surrogate splits:
## H.num.chars < 13.5 to the right, agree=0.764, adj=0.01, (0 split)
## H.num.chars.log < 2.65906 to the right, agree=0.764, adj=0.01, (0 split)
## H.num.words < 11 to the left, agree=0.764, adj=0.01, (0 split)
## H.num.words.unq < 11 to the left, agree=0.764, adj=0.01, (0 split)
## H.num.words.log < 2.481422 to the left, agree=0.764, adj=0.01, (0 split)
##
## Node number 523: 29 observations
## predicted class=N expected loss=0.2413793 P(node) =0.006480447
## class counts: 22 7
## probabilities: 0.759 0.241
##
## Node number 640: 252 observations, complexity param=0.001335113
## predicted class=N expected loss=0.09126984 P(node) =0.05631285
## class counts: 229 23
## probabilities: 0.909 0.091
## left son=1280 (192 obs) right son=1281 (60 obs)
## Primary splits:
## SectionName.fctr < 0.5 to the left, improve=1.8620040, (0 missing)
## PubDate.minute < 5.5 to the left, improve=0.9876580, (0 missing)
## A.num.chars < 101 to the left, improve=0.6567234, (0 missing)
## S.num.chars < 101 to the left, improve=0.6567234, (0 missing)
## S.num.chars.log < 4.624925 to the left, improve=0.6567234, (0 missing)
## Surrogate splits:
## H.num.words < 2.5 to the right, agree=0.889, adj=0.533, (0 split)
## H.num.words.unq < 2.5 to the right, agree=0.889, adj=0.533, (0 split)
## H.num.words.log < 1.242453 to the right, agree=0.889, adj=0.533, (0 split)
## H.num.words.unq.log < 1.242453 to the right, agree=0.889, adj=0.533, (0 split)
## H.num.chars < 23 to the right, agree=0.885, adj=0.517, (0 split)
##
## Node number 641: 15 observations
## predicted class=N expected loss=0.3333333 P(node) =0.003351955
## class counts: 10 5
## probabilities: 0.667 0.333
##
## Node number 642: 94 observations
## predicted class=N expected loss=0.09574468 P(node) =0.02100559
## class counts: 85 9
## probabilities: 0.904 0.096
##
## Node number 643: 462 observations, complexity param=0.001602136
## predicted class=N expected loss=0.2554113 P(node) =0.1032402
## class counts: 344 118
## probabilities: 0.745 0.255
## left son=1286 (39 obs) right son=1287 (423 obs)
## Primary splits:
## SubsectionName.fctrRoom For Debate < 0.5 to the right, improve=4.497630, (0 missing)
## PubDate.hour < 12.5 to the right, improve=4.284205, (0 missing)
## PubDate.wkday < 2.5 to the right, improve=3.982380, (0 missing)
## SectionName.fctrOpinion < 0.5 to the right, improve=3.695456, (0 missing)
## PubDate.second < 53.5 to the left, improve=3.028558, (0 missing)
## Surrogate splits:
## SectionName.fctrOpinion < 0.5 to the right, agree=0.998, adj=0.974, (0 split)
## WordCount < 1909.5 to the right, agree=0.935, adj=0.231, (0 split)
## A.num.chars < 50 to the left, agree=0.918, adj=0.026, (0 split)
## S.num.chars < 50 to the left, agree=0.918, adj=0.026, (0 split)
## S.num.chars.log < 3.931633 to the left, agree=0.918, adj=0.026, (0 split)
##
## Node number 666: 7 observations
## predicted class=N expected loss=0.2857143 P(node) =0.001564246
## class counts: 5 2
## probabilities: 0.714 0.286
##
## Node number 667: 16 observations
## predicted class=Y expected loss=0.125 P(node) =0.003575419
## class counts: 2 14
## probabilities: 0.125 0.875
##
## Node number 962: 12 observations
## predicted class=N expected loss=0.4166667 P(node) =0.002681564
## class counts: 7 5
## probabilities: 0.583 0.417
##
## Node number 963: 19 observations
## predicted class=Y expected loss=0.1578947 P(node) =0.00424581
## class counts: 3 16
## probabilities: 0.158 0.842
##
## Node number 1044: 316 observations
## predicted class=N expected loss=0.04746835 P(node) =0.07061453
## class counts: 301 15
## probabilities: 0.953 0.047
##
## Node number 1045: 99 observations, complexity param=0.0004450378
## predicted class=N expected loss=0.1414141 P(node) =0.02212291
## class counts: 85 14
## probabilities: 0.859 0.141
## left son=2090 (65 obs) right son=2091 (34 obs)
## Primary splits:
## PubDate.month < 9.5 to the right, improve=1.574341, (0 missing)
## PubDate.second < 25.5 to the right, improve=1.247751, (0 missing)
## SectionName.fctrTechnology < 0.5 to the left, improve=0.975469, (0 missing)
## A.num.chars < 175.5 to the right, improve=0.940404, (0 missing)
## S.num.chars < 175.5 to the right, improve=0.940404, (0 missing)
## Surrogate splits:
## H.num.chars < 65.5 to the left, agree=0.687, adj=0.088, (0 split)
## H.num.chars.log < 4.197174 to the left, agree=0.687, adj=0.088, (0 split)
## PubDate.hour < 17.5 to the left, agree=0.667, adj=0.029, (0 split)
## PubDate.date < 26.5 to the right, agree=0.667, adj=0.029, (0 split)
## A.num.chars < 72.5 to the right, agree=0.667, adj=0.029, (0 split)
##
## Node number 1280: 192 observations
## predicted class=N expected loss=0.05729167 P(node) =0.04290503
## class counts: 181 11
## probabilities: 0.943 0.057
##
## Node number 1281: 60 observations, complexity param=0.001335113
## predicted class=N expected loss=0.2 P(node) =0.01340782
## class counts: 48 12
## probabilities: 0.800 0.200
## left son=2562 (31 obs) right son=2563 (29 obs)
## Primary splits:
## WordCount < 1184.5 to the right, improve=5.131034, (0 missing)
## PubDate.hour < 6.5 to the right, improve=5.131034, (0 missing)
## H.num.chars < 18 to the left, improve=4.490323, (0 missing)
## H.num.chars.log < 2.943052 to the left, improve=4.490323, (0 missing)
## H.num.words < 2.5 to the left, improve=1.745455, (0 missing)
## Surrogate splits:
## H.num.chars < 18 to the left, agree=0.867, adj=0.724, (0 split)
## H.num.chars.log < 2.943052 to the left, agree=0.867, adj=0.724, (0 split)
## PubDate.hour < 6.5 to the right, agree=0.833, adj=0.655, (0 split)
## H.num.words < 2.5 to the left, agree=0.833, adj=0.655, (0 split)
## H.num.words.unq < 2.5 to the left, agree=0.833, adj=0.655, (0 split)
##
## Node number 1286: 39 observations
## predicted class=N expected loss=0.02564103 P(node) =0.008715084
## class counts: 38 1
## probabilities: 0.974 0.026
##
## Node number 1287: 423 observations, complexity param=0.001602136
## predicted class=N expected loss=0.2765957 P(node) =0.09452514
## class counts: 306 117
## probabilities: 0.723 0.277
## left son=2574 (287 obs) right son=2575 (136 obs)
## Primary splits:
## WordCount < 1152.5 to the left, improve=4.483810, (0 missing)
## PubDate.wkday < 2.5 to the right, improve=4.393720, (0 missing)
## NewsDesk.fctrTStyle < 0.5 to the right, improve=3.860776, (0 missing)
## H.num.words.unq < 4.5 to the right, improve=3.708937, (0 missing)
## H.num.words.unq.log < 1.700599 to the right, improve=3.708937, (0 missing)
## Surrogate splits:
## PubDate.hour < 8.5 to the right, agree=0.704, adj=0.081, (0 split)
## H.num.chars < 25.5 to the right, agree=0.702, adj=0.074, (0 split)
## H.num.chars.log < 3.276967 to the right, agree=0.702, adj=0.074, (0 split)
## PubDate.minute < 1.5 to the right, agree=0.697, adj=0.059, (0 split)
## A.num.chars < 186.5 to the left, agree=0.686, adj=0.022, (0 split)
##
## Node number 2090: 65 observations
## predicted class=N expected loss=0.07692308 P(node) =0.01452514
## class counts: 60 5
## probabilities: 0.923 0.077
##
## Node number 2091: 34 observations, complexity param=0.0004450378
## predicted class=N expected loss=0.2647059 P(node) =0.007597765
## class counts: 25 9
## probabilities: 0.735 0.265
## left son=4182 (8 obs) right son=4183 (26 obs)
## Primary splits:
## A.num.words < 16.5 to the right, improve=1.466063, (0 missing)
## S.num.words < 16.5 to the right, improve=1.466063, (0 missing)
## A.num.words.log < 2.861793 to the right, improve=1.466063, (0 missing)
## S.num.words.log < 2.861793 to the right, improve=1.466063, (0 missing)
## NewsDesk.fctrTStyle < 0.5 to the right, improve=1.235294, (0 missing)
## Surrogate splits:
## S.num.words < 16.5 to the right, agree=1.000, adj=1.000, (0 split)
## A.num.words.log < 2.861793 to the right, agree=1.000, adj=1.000, (0 split)
## S.num.words.log < 2.861793 to the right, agree=1.000, adj=1.000, (0 split)
## A.num.words.unq < 16.5 to the right, agree=0.971, adj=0.875, (0 split)
## S.num.words.unq < 16.5 to the right, agree=0.971, adj=0.875, (0 split)
##
## Node number 2562: 31 observations
## predicted class=N expected loss=0 P(node) =0.006927374
## class counts: 31 0
## probabilities: 1.000 0.000
##
## Node number 2563: 29 observations, complexity param=0.001335113
## predicted class=N expected loss=0.4137931 P(node) =0.006480447
## class counts: 17 12
## probabilities: 0.586 0.414
## left son=5126 (19 obs) right son=5127 (10 obs)
## Primary splits:
## PubDate.wkday < 3.5 to the left, improve=2.5005440, (0 missing)
## PubDate.minute < 26 to the right, improve=1.8427750, (0 missing)
## PubDate.second < 23.5 to the right, improve=1.0584390, (0 missing)
## WordCount < 900 to the left, improve=0.7053292, (0 missing)
## PubDate.date < 9.5 to the left, improve=0.7053292, (0 missing)
## Surrogate splits:
## PubDate.date < 15.5 to the left, agree=0.724, adj=0.2, (0 split)
## S.new < 0.5 to the left, agree=0.724, adj=0.2, (0 split)
## A.new < 0.5 to the left, agree=0.724, adj=0.2, (0 split)
## A.num.chars < 200.5 to the left, agree=0.724, adj=0.2, (0 split)
## S.num.chars < 200.5 to the left, agree=0.724, adj=0.2, (0 split)
##
## Node number 2574: 287 observations, complexity param=0.001602136
## predicted class=N expected loss=0.2264808 P(node) =0.06413408
## class counts: 222 65
## probabilities: 0.774 0.226
## left son=5148 (214 obs) right son=5149 (73 obs)
## Primary splits:
## PubDate.wkday < 1.5 to the right, improve=4.025421, (0 missing)
## PubDate.second < 54.5 to the left, improve=3.159319, (0 missing)
## NewsDesk.fctr < 0.5 to the left, improve=2.540345, (0 missing)
## NewsDesk.fctrTStyle < 0.5 to the right, improve=2.378395, (0 missing)
## H.num.words.unq < 4.5 to the right, improve=2.280619, (0 missing)
## Surrogate splits:
## PubDate.hour < 20.5 to the left, agree=0.770, adj=0.096, (0 split)
## H.num.chars < 14 to the right, agree=0.753, adj=0.027, (0 split)
## H.num.chars.log < 2.705823 to the right, agree=0.753, adj=0.027, (0 split)
## A.num.chars < 47 to the right, agree=0.753, adj=0.027, (0 split)
## S.num.chars < 47 to the right, agree=0.753, adj=0.027, (0 split)
##
## Node number 2575: 136 observations, complexity param=0.001602136
## predicted class=N expected loss=0.3823529 P(node) =0.03039106
## class counts: 84 52
## probabilities: 0.618 0.382
## left son=5150 (126 obs) right son=5151 (10 obs)
## Primary splits:
## H.num.chars < 21.5 to the right, improve=3.765453, (0 missing)
## H.num.chars.log < 3.113268 to the right, improve=3.765453, (0 missing)
## PubDate.wkday < 3.5 to the right, improve=3.432863, (0 missing)
## NewsDesk.fctrTStyle < 0.5 to the right, improve=2.405360, (0 missing)
## PubDate.hour < 8.5 to the right, improve=2.171264, (0 missing)
## Surrogate splits:
## H.num.chars.log < 3.113268 to the right, agree=1.000, adj=1.0, (0 split)
## H.num.words < 1.5 to the right, agree=0.941, adj=0.2, (0 split)
## H.num.words.unq < 1.5 to the right, agree=0.941, adj=0.2, (0 split)
## H.num.words.log < 0.8958797 to the right, agree=0.941, adj=0.2, (0 split)
## H.num.words.unq.log < 0.8958797 to the right, agree=0.941, adj=0.2, (0 split)
##
## Node number 4182: 8 observations
## predicted class=N expected loss=0 P(node) =0.001787709
## class counts: 8 0
## probabilities: 1.000 0.000
##
## Node number 4183: 26 observations, complexity param=0.0004450378
## predicted class=N expected loss=0.3461538 P(node) =0.005810056
## class counts: 17 9
## probabilities: 0.654 0.346
## left son=8366 (18 obs) right son=8367 (8 obs)
## Primary splits:
## SectionName.fctrTechnology < 0.5 to the left, improve=1.797009, (0 missing)
## A.num.words < 13.5 to the left, improve=1.207139, (0 missing)
## S.num.words < 13.5 to the left, improve=1.207139, (0 missing)
## A.num.words.log < 2.673554 to the left, improve=1.207139, (0 missing)
## S.num.words.log < 2.673554 to the left, improve=1.207139, (0 missing)
## Surrogate splits:
## WordCount < 591.5 to the left, agree=0.769, adj=0.25, (0 split)
## PubDate.minute < 57 to the left, agree=0.769, adj=0.25, (0 split)
## A.num.chars < 161 to the left, agree=0.769, adj=0.25, (0 split)
## S.num.chars < 161 to the left, agree=0.769, adj=0.25, (0 split)
## S.num.chars.log < 5.085687 to the left, agree=0.769, adj=0.25, (0 split)
##
## Node number 5126: 19 observations
## predicted class=N expected loss=0.2631579 P(node) =0.00424581
## class counts: 14 5
## probabilities: 0.737 0.263
##
## Node number 5127: 10 observations
## predicted class=Y expected loss=0.3 P(node) =0.002234637
## class counts: 3 7
## probabilities: 0.300 0.700
##
## Node number 5148: 214 observations, complexity param=0.001602136
## predicted class=N expected loss=0.1775701 P(node) =0.04782123
## class counts: 176 38
## probabilities: 0.822 0.178
## left son=10296 (192 obs) right son=10297 (22 obs)
## Primary splits:
## PubDate.second < 53.5 to the left, improve=2.628726, (0 missing)
## PubDate.hour < 12.5 to the right, improve=1.833210, (0 missing)
## NewsDesk.fctr < 0.5 to the left, improve=1.525789, (0 missing)
## WordCount < 992 to the left, improve=1.427800, (0 missing)
## PubDate.date < 24.5 to the left, improve=1.098137, (0 missing)
## Surrogate splits:
## A.num.chars < 186 to the left, agree=0.907, adj=0.091, (0 split)
## S.num.chars < 186 to the left, agree=0.907, adj=0.091, (0 split)
## S.num.chars.log < 5.231094 to the left, agree=0.907, adj=0.091, (0 split)
## A.num.chars.log < 5.231094 to the left, agree=0.907, adj=0.091, (0 split)
## PubDate.minute < 56.5 to the left, agree=0.902, adj=0.045, (0 split)
##
## Node number 5149: 73 observations, complexity param=0.001602136
## predicted class=N expected loss=0.369863 P(node) =0.01631285
## class counts: 46 27
## probabilities: 0.630 0.370
## left son=10298 (50 obs) right son=10299 (23 obs)
## Primary splits:
## H.num.words.unq < 4.5 to the right, improve=3.830876, (0 missing)
## H.num.words.unq.log < 1.700599 to the right, improve=3.830876, (0 missing)
## H.num.words < 4.5 to the right, improve=3.077308, (0 missing)
## H.num.words.log < 1.700599 to the right, improve=3.077308, (0 missing)
## H.num.chars < 31.5 to the right, improve=2.596628, (0 missing)
## Surrogate splits:
## H.num.words.unq.log < 1.700599 to the right, agree=1.000, adj=1.000, (0 split)
## H.num.words < 4.5 to the right, agree=0.986, adj=0.957, (0 split)
## H.num.words.log < 1.700599 to the right, agree=0.986, adj=0.957, (0 split)
## H.num.chars < 44.5 to the right, agree=0.918, adj=0.739, (0 split)
## H.num.chars.log < 3.817652 to the right, agree=0.918, adj=0.739, (0 split)
##
## Node number 5150: 126 observations, complexity param=0.001602136
## predicted class=N expected loss=0.3492063 P(node) =0.02815642
## class counts: 82 44
## probabilities: 0.651 0.349
## left son=10300 (84 obs) right son=10301 (42 obs)
## Primary splits:
## PubDate.second < 20.5 to the right, improve=2.865079, (0 missing)
## PubDate.wkday < 3.5 to the right, improve=2.293651, (0 missing)
## PubDate.date < 29.5 to the right, improve=2.083401, (0 missing)
## PubDate.hour < 20.5 to the left, improve=1.953602, (0 missing)
## A.num.words.unq < 16.5 to the right, improve=1.848966, (0 missing)
## Surrogate splits:
## PubDate.minute < 51.5 to the left, agree=0.690, adj=0.071, (0 split)
## WordCount < 1156.5 to the right, agree=0.683, adj=0.048, (0 split)
## H.num.words.unq < 7.5 to the left, agree=0.683, adj=0.048, (0 split)
## A.num.words < 20.5 to the left, agree=0.683, adj=0.048, (0 split)
## S.num.words < 20.5 to the left, agree=0.683, adj=0.048, (0 split)
##
## Node number 5151: 10 observations
## predicted class=Y expected loss=0.2 P(node) =0.002234637
## class counts: 2 8
## probabilities: 0.200 0.800
##
## Node number 8366: 18 observations
## predicted class=N expected loss=0.2222222 P(node) =0.004022346
## class counts: 14 4
## probabilities: 0.778 0.222
##
## Node number 8367: 8 observations
## predicted class=Y expected loss=0.375 P(node) =0.001787709
## class counts: 3 5
## probabilities: 0.375 0.625
##
## Node number 10296: 192 observations, complexity param=0.001335113
## predicted class=N expected loss=0.1510417 P(node) =0.04290503
## class counts: 163 29
## probabilities: 0.849 0.151
## left son=20592 (158 obs) right son=20593 (34 obs)
## Primary splits:
## PubDate.date < 24.5 to the left, improve=1.6915570, (0 missing)
## WordCount < 992 to the left, improve=1.6515280, (0 missing)
## A.num.words < 17.5 to the right, improve=0.9621267, (0 missing)
## S.num.words < 17.5 to the right, improve=0.9621267, (0 missing)
## A.num.words.log < 2.917405 to the right, improve=0.9621267, (0 missing)
## Surrogate splits:
## A.num.chars < 184.5 to the left, agree=0.839, adj=0.088, (0 split)
## S.num.chars < 184.5 to the left, agree=0.839, adj=0.088, (0 split)
## S.num.chars.log < 5.223051 to the left, agree=0.839, adj=0.088, (0 split)
## A.num.chars.log < 5.223051 to the left, agree=0.839, adj=0.088, (0 split)
## H.num.chars < 90 to the left, agree=0.828, adj=0.029, (0 split)
##
## Node number 10297: 22 observations, complexity param=0.001602136
## predicted class=N expected loss=0.4090909 P(node) =0.004916201
## class counts: 13 9
## probabilities: 0.591 0.409
## left son=20594 (12 obs) right son=20595 (10 obs)
## Primary splits:
## PubDate.hour < 12.5 to the right, improve=3.1030300, (0 missing)
## PubDate.second < 57.5 to the right, improve=2.0292210, (0 missing)
## PubDate.minute < 28 to the right, improve=1.3363640, (0 missing)
## WordCount < 844 to the left, improve=0.8181818, (0 missing)
## PubDate.wkday < 3.5 to the left, improve=0.8181818, (0 missing)
## Surrogate splits:
## A.num.chars < 163 to the left, agree=0.727, adj=0.4, (0 split)
## S.num.chars < 163 to the left, agree=0.727, adj=0.4, (0 split)
## S.num.chars.log < 5.099848 to the left, agree=0.727, adj=0.4, (0 split)
## A.num.chars.log < 5.099848 to the left, agree=0.727, adj=0.4, (0 split)
## PubDate.minute < 25.5 to the right, agree=0.682, adj=0.3, (0 split)
##
## Node number 10298: 50 observations, complexity param=0.001335113
## predicted class=N expected loss=0.26 P(node) =0.01117318
## class counts: 37 13
## probabilities: 0.740 0.260
## left son=20596 (39 obs) right son=20597 (11 obs)
## Primary splits:
## H.num.words < 7.5 to the left, improve=2.298275, (0 missing)
## H.num.words.unq < 7.5 to the left, improve=2.298275, (0 missing)
## H.num.words.log < 2.138333 to the left, improve=2.298275, (0 missing)
## H.num.words.unq.log < 2.138333 to the left, improve=2.298275, (0 missing)
## WordCount < 988 to the right, improve=2.084920, (0 missing)
## Surrogate splits:
## H.num.words.unq < 7.5 to the left, agree=1.00, adj=1.000, (0 split)
## H.num.words.log < 2.138333 to the left, agree=1.00, adj=1.000, (0 split)
## H.num.words.unq.log < 2.138333 to the left, agree=1.00, adj=1.000, (0 split)
## H.num.chars < 63.5 to the left, agree=0.84, adj=0.273, (0 split)
## H.num.chars.log < 4.166635 to the left, agree=0.84, adj=0.273, (0 split)
##
## Node number 10299: 23 observations
## predicted class=Y expected loss=0.3913043 P(node) =0.005139665
## class counts: 9 14
## probabilities: 0.391 0.609
##
## Node number 10300: 84 observations, complexity param=0.001602136
## predicted class=N expected loss=0.2738095 P(node) =0.01877095
## class counts: 61 23
## probabilities: 0.726 0.274
## left son=20600 (62 obs) right son=20601 (22 obs)
## Primary splits:
## PubDate.month < 10.5 to the left, improve=3.049923, (0 missing)
## PubDate.hour < 20.5 to the left, improve=2.963203, (0 missing)
## NewsDesk.fctrTStyle < 0.5 to the right, improve=2.738095, (0 missing)
## SectionName.fctrBusiness Day < 0.5 to the left, improve=2.460020, (0 missing)
## SubsectionName.fctrDealbook < 0.5 to the left, improve=2.460020, (0 missing)
## Surrogate splits:
## PubDate.second < 53.5 to the left, agree=0.762, adj=0.091, (0 split)
## H.new < 0.5 to the left, agree=0.762, adj=0.091, (0 split)
##
## Node number 10301: 42 observations, complexity param=0.001602136
## predicted class=N expected loss=0.5 P(node) =0.009385475
## class counts: 21 21
## probabilities: 0.500 0.500
## left son=20602 (25 obs) right son=20603 (17 obs)
## Primary splits:
## A.num.chars < 128.5 to the right, improve=4.002353, (0 missing)
## S.num.chars < 128.5 to the right, improve=4.002353, (0 missing)
## S.num.chars.log < 4.863673 to the right, improve=4.002353, (0 missing)
## A.num.chars.log < 4.863673 to the right, improve=4.002353, (0 missing)
## A.num.words < 12.5 to the right, improve=3.428571, (0 missing)
## Surrogate splits:
## S.num.chars < 128.5 to the right, agree=1.000, adj=1.000, (0 split)
## S.num.chars.log < 4.863673 to the right, agree=1.000, adj=1.000, (0 split)
## A.num.chars.log < 4.863673 to the right, agree=1.000, adj=1.000, (0 split)
## A.num.words < 12.5 to the right, agree=0.929, adj=0.824, (0 split)
## S.num.words < 12.5 to the right, agree=0.929, adj=0.824, (0 split)
##
## Node number 20592: 158 observations, complexity param=0.001335113
## predicted class=N expected loss=0.1202532 P(node) =0.03530726
## class counts: 139 19
## probabilities: 0.880 0.120
## left son=41184 (116 obs) right son=41185 (42 obs)
## Primary splits:
## WordCount < 992 to the left, improve=1.5888360, (0 missing)
## PubDate.minute < 17.5 to the left, improve=0.8515919, (0 missing)
## PubDate.hour < 12.5 to the right, improve=0.7468218, (0 missing)
## NewsDesk.fctr < 0.5 to the left, improve=0.7468218, (0 missing)
## S.new < 0.5 to the right, improve=0.6246244, (0 missing)
## Surrogate splits:
## H.num.chars < 27.5 to the right, agree=0.753, adj=0.071, (0 split)
## H.num.chars.log < 3.34975 to the right, agree=0.753, adj=0.071, (0 split)
##
## Node number 20593: 34 observations, complexity param=0.001335113
## predicted class=N expected loss=0.2941176 P(node) =0.007597765
## class counts: 24 10
## probabilities: 0.706 0.294
## left son=41186 (24 obs) right son=41187 (10 obs)
## Primary splits:
## PubDate.wkday < 2.5 to the right, improve=2.650980, (0 missing)
## A.num.words < 14.5 to the right, improve=2.360504, (0 missing)
## S.num.words < 14.5 to the right, improve=2.360504, (0 missing)
## A.num.words.log < 2.740319 to the right, improve=2.360504, (0 missing)
## S.num.words.log < 2.740319 to the right, improve=2.360504, (0 missing)
## Surrogate splits:
## PubDate.hour < 17.5 to the left, agree=0.765, adj=0.2, (0 split)
## A.num.chars < 81.5 to the right, agree=0.735, adj=0.1, (0 split)
## S.num.chars < 81.5 to the right, agree=0.735, adj=0.1, (0 split)
## H.num.words < 9.5 to the left, agree=0.735, adj=0.1, (0 split)
## H.num.words.unq < 9.5 to the left, agree=0.735, adj=0.1, (0 split)
##
## Node number 20594: 12 observations
## predicted class=N expected loss=0.1666667 P(node) =0.002681564
## class counts: 10 2
## probabilities: 0.833 0.167
##
## Node number 20595: 10 observations
## predicted class=Y expected loss=0.3 P(node) =0.002234637
## class counts: 3 7
## probabilities: 0.300 0.700
##
## Node number 20596: 39 observations
## predicted class=N expected loss=0.1794872 P(node) =0.008715084
## class counts: 32 7
## probabilities: 0.821 0.179
##
## Node number 20597: 11 observations
## predicted class=Y expected loss=0.4545455 P(node) =0.002458101
## class counts: 5 6
## probabilities: 0.455 0.545
##
## Node number 20600: 62 observations, complexity param=0.001602136
## predicted class=N expected loss=0.1935484 P(node) =0.01385475
## class counts: 50 12
## probabilities: 0.806 0.194
## left son=41200 (28 obs) right son=41201 (34 obs)
## Primary splits:
## PubDate.month < 9.5 to the right, improve=2.543914, (0 missing)
## A.num.chars < 178.5 to the left, improve=2.253540, (0 missing)
## S.num.chars < 178.5 to the left, improve=2.253540, (0 missing)
## S.num.chars.log < 5.190171 to the left, improve=2.253540, (0 missing)
## A.num.chars.log < 5.190171 to the left, improve=2.253540, (0 missing)
## Surrogate splits:
## PubDate.date < 26 to the right, agree=0.677, adj=0.286, (0 split)
## WordCount < 1863 to the right, agree=0.661, adj=0.250, (0 split)
## PubDate.hour < 12.5 to the left, agree=0.661, adj=0.250, (0 split)
## NewsDesk.fctrTStyle < 0.5 to the right, agree=0.645, adj=0.214, (0 split)
## PubDate.second < 34.5 to the left, agree=0.613, adj=0.143, (0 split)
##
## Node number 20601: 22 observations, complexity param=0.001602136
## predicted class=N expected loss=0.5 P(node) =0.004916201
## class counts: 11 11
## probabilities: 0.500 0.500
## left son=41202 (9 obs) right son=41203 (13 obs)
## Primary splits:
## PubDate.date < 10.5 to the left, improve=2.350427, (0 missing)
## H.num.words < 6.5 to the left, improve=1.571429, (0 missing)
## H.num.words.unq < 6.5 to the left, improve=1.571429, (0 missing)
## H.num.words.log < 2.012676 to the left, improve=1.571429, (0 missing)
## A.num.words < 15.5 to the right, improve=1.571429, (0 missing)
## Surrogate splits:
## A.num.words < 15.5 to the right, agree=0.773, adj=0.444, (0 split)
## S.num.words < 15.5 to the right, agree=0.773, adj=0.444, (0 split)
## A.num.words.unq < 15.5 to the right, agree=0.773, adj=0.444, (0 split)
## S.num.words.unq < 15.5 to the right, agree=0.773, adj=0.444, (0 split)
## A.num.words.log < 2.802901 to the right, agree=0.773, adj=0.444, (0 split)
##
## Node number 20602: 25 observations, complexity param=0.001602136
## predicted class=N expected loss=0.32 P(node) =0.005586592
## class counts: 17 8
## probabilities: 0.680 0.320
## left son=41204 (17 obs) right son=41205 (8 obs)
## Primary splits:
## PubDate.month < 10.5 to the left, improve=2.188824, (0 missing)
## PubDate.minute < 4.5 to the right, improve=1.560556, (0 missing)
## H.num.chars < 64.5 to the left, improve=1.229206, (0 missing)
## H.num.chars.log < 4.182021 to the left, improve=1.229206, (0 missing)
## PubDate.wkday < 3.5 to the right, improve=1.227222, (0 missing)
## Surrogate splits:
## WordCount < 1166 to the right, agree=0.76, adj=0.250, (0 split)
## PubDate.hour < 10 to the right, agree=0.76, adj=0.250, (0 split)
## H.num.chars < 71.5 to the left, agree=0.76, adj=0.250, (0 split)
## H.num.chars.log < 4.283563 to the left, agree=0.76, adj=0.250, (0 split)
## NewsDesk.fctrTStyle < 0.5 to the left, agree=0.72, adj=0.125, (0 split)
##
## Node number 20603: 17 observations
## predicted class=Y expected loss=0.2352941 P(node) =0.003798883
## class counts: 4 13
## probabilities: 0.235 0.765
##
## Node number 41184: 116 observations
## predicted class=N expected loss=0.07758621 P(node) =0.02592179
## class counts: 107 9
## probabilities: 0.922 0.078
##
## Node number 41185: 42 observations, complexity param=0.001335113
## predicted class=N expected loss=0.2380952 P(node) =0.009385475
## class counts: 32 10
## probabilities: 0.762 0.238
## left son=82370 (21 obs) right son=82371 (21 obs)
## Primary splits:
## WordCount < 1073.5 to the right, improve=3.047619, (0 missing)
## PubDate.second < 44.5 to the left, improve=1.866667, (0 missing)
## PubDate.wkday < 2.5 to the left, improve=1.689708, (0 missing)
## PubDate.minute < 43.5 to the left, improve=1.355742, (0 missing)
## H.num.words < 3.5 to the left, improve=1.298701, (0 missing)
## Surrogate splits:
## PubDate.second < 31 to the left, agree=0.69, adj=0.381, (0 split)
## A.num.chars < 155.5 to the left, agree=0.69, adj=0.381, (0 split)
## S.num.chars < 155.5 to the left, agree=0.69, adj=0.381, (0 split)
## S.num.chars.log < 5.05301 to the left, agree=0.69, adj=0.381, (0 split)
## A.num.chars.log < 5.05301 to the left, agree=0.69, adj=0.381, (0 split)
##
## Node number 41186: 24 observations
## predicted class=N expected loss=0.1666667 P(node) =0.005363128
## class counts: 20 4
## probabilities: 0.833 0.167
##
## Node number 41187: 10 observations
## predicted class=Y expected loss=0.4 P(node) =0.002234637
## class counts: 4 6
## probabilities: 0.400 0.600
##
## Node number 41200: 28 observations
## predicted class=N expected loss=0.03571429 P(node) =0.006256983
## class counts: 27 1
## probabilities: 0.964 0.036
##
## Node number 41201: 34 observations, complexity param=0.001602136
## predicted class=N expected loss=0.3235294 P(node) =0.007597765
## class counts: 23 11
## probabilities: 0.676 0.324
## left son=82402 (27 obs) right son=82403 (7 obs)
## Primary splits:
## A.num.chars < 172 to the left, improve=2.691877, (0 missing)
## S.num.chars < 172 to the left, improve=2.691877, (0 missing)
## S.num.chars.log < 5.153024 to the left, improve=2.691877, (0 missing)
## A.num.chars.log < 5.153024 to the left, improve=2.691877, (0 missing)
## PubDate.date < 9.5 to the right, improve=1.901584, (0 missing)
## Surrogate splits:
## S.num.chars < 172 to the left, agree=1.000, adj=1.000, (0 split)
## S.num.chars.log < 5.153024 to the left, agree=1.000, adj=1.000, (0 split)
## A.num.chars.log < 5.153024 to the left, agree=1.000, adj=1.000, (0 split)
## A.num.words < 14.5 to the left, agree=0.853, adj=0.286, (0 split)
## S.num.words < 14.5 to the left, agree=0.853, adj=0.286, (0 split)
##
## Node number 41202: 9 observations
## predicted class=N expected loss=0.2222222 P(node) =0.002011173
## class counts: 7 2
## probabilities: 0.778 0.222
##
## Node number 41203: 13 observations
## predicted class=Y expected loss=0.3076923 P(node) =0.002905028
## class counts: 4 9
## probabilities: 0.308 0.692
##
## Node number 41204: 17 observations
## predicted class=N expected loss=0.1764706 P(node) =0.003798883
## class counts: 14 3
## probabilities: 0.824 0.176
##
## Node number 41205: 8 observations
## predicted class=Y expected loss=0.375 P(node) =0.001787709
## class counts: 3 5
## probabilities: 0.375 0.625
##
## Node number 82370: 21 observations
## predicted class=N expected loss=0.04761905 P(node) =0.004692737
## class counts: 20 1
## probabilities: 0.952 0.048
##
## Node number 82371: 21 observations, complexity param=0.001335113
## predicted class=N expected loss=0.4285714 P(node) =0.004692737
## class counts: 12 9
## probabilities: 0.571 0.429
## left son=164742 (14 obs) right son=164743 (7 obs)
## Primary splits:
## PubDate.minute < 34 to the left, improve=1.714286, (0 missing)
## A.num.chars < 140.5 to the right, improve=1.714286, (0 missing)
## S.num.chars < 140.5 to the right, improve=1.714286, (0 missing)
## S.num.chars.log < 4.952244 to the right, improve=1.714286, (0 missing)
## A.num.chars.log < 4.952244 to the right, improve=1.714286, (0 missing)
## Surrogate splits:
## WordCount < 1041 to the left, agree=0.810, adj=0.429, (0 split)
## PubDate.date < 3.5 to the right, agree=0.810, adj=0.429, (0 split)
## SectionName.fctrBusiness Day < 0.5 to the left, agree=0.810, adj=0.429, (0 split)
## SubsectionName.fctrDealbook < 0.5 to the left, agree=0.810, adj=0.429, (0 split)
## PubDate.hour < 18.5 to the left, agree=0.714, adj=0.143, (0 split)
##
## Node number 82402: 27 observations
## predicted class=N expected loss=0.2222222 P(node) =0.00603352
## class counts: 21 6
## probabilities: 0.778 0.222
##
## Node number 82403: 7 observations
## predicted class=Y expected loss=0.2857143 P(node) =0.001564246
## class counts: 2 5
## probabilities: 0.286 0.714
##
## Node number 164742: 14 observations
## predicted class=N expected loss=0.2857143 P(node) =0.003128492
## class counts: 10 4
## probabilities: 0.714 0.286
##
## Node number 164743: 7 observations
## predicted class=Y expected loss=0.2857143 P(node) =0.001564246
## class counts: 2 5
## probabilities: 0.286 0.714
##
## n= 4475
##
## node), split, n, loss, yval, (yprob)
## * denotes terminal node
##
## 1) root 4475 749 N (0.83262570 0.16737430)
## 2) NewsDesk.fctrOpEd< 0.5 4131 480 N (0.88380537 0.11619463)
## 4) WordCount< 683.5 3066 152 N (0.95042401 0.04957599)
## 8) PubDate.hour< 21.5 2987 118 N (0.96049548 0.03950452)
## 16) NewsDesk.fctrStyles< 0.5 2864 86 N (0.96997207 0.03002793)
## 32) NewsDesk.fctrScience< 0.5 2806 69 N (0.97540984 0.02459016)
## 64) WordCount< 396.5 2102 24 N (0.98858230 0.01141770) *
## 65) WordCount>=396.5 704 45 N (0.93607955 0.06392045)
## 130) A.num.chars>=56.5 694 37 N (0.94668588 0.05331412)
## 260) SectionName.fctrBusiness Day>=0.5 250 1 N (0.99600000 0.00400000) *
## 261) SectionName.fctrBusiness Day< 0.5 444 36 N (0.91891892 0.08108108)
## 522) PubDate.hour< 18.5 415 29 N (0.93012048 0.06987952)
## 1044) PubDate.date< 23.5 316 15 N (0.95253165 0.04746835) *
## 1045) PubDate.date>=23.5 99 14 N (0.85858586 0.14141414)
## 2090) PubDate.month>=9.5 65 5 N (0.92307692 0.07692308) *
## 2091) PubDate.month< 9.5 34 9 N (0.73529412 0.26470588)
## 4182) A.num.words>=16.5 8 0 N (1.00000000 0.00000000) *
## 4183) A.num.words< 16.5 26 9 N (0.65384615 0.34615385)
## 8366) SectionName.fctrTechnology< 0.5 18 4 N (0.77777778 0.22222222) *
## 8367) SectionName.fctrTechnology>=0.5 8 3 Y (0.37500000 0.62500000) *
## 523) PubDate.hour>=18.5 29 7 N (0.75862069 0.24137931) *
## 131) A.num.chars< 56.5 10 2 Y (0.20000000 0.80000000) *
## 33) NewsDesk.fctrScience>=0.5 58 17 N (0.70689655 0.29310345)
## 66) PubDate.hour>=9.5 48 10 N (0.79166667 0.20833333) *
## 67) PubDate.hour< 9.5 10 3 Y (0.30000000 0.70000000) *
## 17) NewsDesk.fctrStyles>=0.5 123 32 N (0.73983740 0.26016260)
## 34) SectionName.fctrU.S.< 0.5 61 0 N (1.00000000 0.00000000) *
## 35) SectionName.fctrU.S.>=0.5 62 30 Y (0.48387097 0.51612903)
## 70) PubDate.second>=8.5 53 24 N (0.54716981 0.45283019)
## 140) S.week>=0.5 16 3 N (0.81250000 0.18750000) *
## 141) S.week< 0.5 37 16 Y (0.43243243 0.56756757)
## 282) PubDate.second< 21.5 9 2 N (0.77777778 0.22222222) *
## 283) PubDate.second>=21.5 28 9 Y (0.32142857 0.67857143) *
## 71) PubDate.second< 8.5 9 1 Y (0.11111111 0.88888889) *
## 9) PubDate.hour>=21.5 79 34 N (0.56962025 0.43037975)
## 18) PubDate.minute>=0.5 45 5 N (0.88888889 0.11111111) *
## 19) PubDate.minute< 0.5 34 5 Y (0.14705882 0.85294118)
## 38) WordCount< 327.5 7 3 N (0.57142857 0.42857143) *
## 39) WordCount>=327.5 27 1 Y (0.03703704 0.96296296) *
## 5) WordCount>=683.5 1065 328 N (0.69201878 0.30798122)
## 10) SectionName.fctrHealth< 0.5 994 263 N (0.73541247 0.26458753)
## 20) PubDate.hour< 21.5 929 220 N (0.76318622 0.23681378)
## 40) NewsDesk.fctrStyles< 0.5 851 177 N (0.79200940 0.20799060)
## 80) PubDate.wkday< 5.5 834 164 N (0.80335731 0.19664269)
## 160) SubsectionName.fctrThe Public Editor< 0.5 823 155 N (0.81166464 0.18833536)
## 320) PubDate.hour< 7.5 267 28 N (0.89513109 0.10486891)
## 640) S.compani< 0.5 252 23 N (0.90873016 0.09126984)
## 1280) SectionName.fctr< 0.5 192 11 N (0.94270833 0.05729167) *
## 1281) SectionName.fctr>=0.5 60 12 N (0.80000000 0.20000000)
## 2562) WordCount>=1184.5 31 0 N (1.00000000 0.00000000) *
## 2563) WordCount< 1184.5 29 12 N (0.58620690 0.41379310)
## 5126) PubDate.wkday< 3.5 19 5 N (0.73684211 0.26315789) *
## 5127) PubDate.wkday>=3.5 10 3 Y (0.30000000 0.70000000) *
## 641) S.compani>=0.5 15 5 N (0.66666667 0.33333333) *
## 321) PubDate.hour>=7.5 556 127 N (0.77158273 0.22841727)
## 642) A.num.chars>=188.5 94 9 N (0.90425532 0.09574468) *
## 643) A.num.chars< 188.5 462 118 N (0.74458874 0.25541126)
## 1286) SubsectionName.fctrRoom For Debate>=0.5 39 1 N (0.97435897 0.02564103) *
## 1287) SubsectionName.fctrRoom For Debate< 0.5 423 117 N (0.72340426 0.27659574)
## 2574) WordCount< 1152.5 287 65 N (0.77351916 0.22648084)
## 5148) PubDate.wkday>=1.5 214 38 N (0.82242991 0.17757009)
## 10296) PubDate.second< 53.5 192 29 N (0.84895833 0.15104167)
## 20592) PubDate.date< 24.5 158 19 N (0.87974684 0.12025316)
## 41184) WordCount< 992 116 9 N (0.92241379 0.07758621) *
## 41185) WordCount>=992 42 10 N (0.76190476 0.23809524)
## 82370) WordCount>=1073.5 21 1 N (0.95238095 0.04761905) *
## 82371) WordCount< 1073.5 21 9 N (0.57142857 0.42857143)
## 164742) PubDate.minute< 34 14 4 N (0.71428571 0.28571429) *
## 164743) PubDate.minute>=34 7 2 Y (0.28571429 0.71428571) *
## 20593) PubDate.date>=24.5 34 10 N (0.70588235 0.29411765)
## 41186) PubDate.wkday>=2.5 24 4 N (0.83333333 0.16666667) *
## 41187) PubDate.wkday< 2.5 10 4 Y (0.40000000 0.60000000) *
## 10297) PubDate.second>=53.5 22 9 N (0.59090909 0.40909091)
## 20594) PubDate.hour>=12.5 12 2 N (0.83333333 0.16666667) *
## 20595) PubDate.hour< 12.5 10 3 Y (0.30000000 0.70000000) *
## 5149) PubDate.wkday< 1.5 73 27 N (0.63013699 0.36986301)
## 10298) H.num.words.unq>=4.5 50 13 N (0.74000000 0.26000000)
## 20596) H.num.words< 7.5 39 7 N (0.82051282 0.17948718) *
## 20597) H.num.words>=7.5 11 5 Y (0.45454545 0.54545455) *
## 10299) H.num.words.unq< 4.5 23 9 Y (0.39130435 0.60869565) *
## 2575) WordCount>=1152.5 136 52 N (0.61764706 0.38235294)
## 5150) H.num.chars>=21.5 126 44 N (0.65079365 0.34920635)
## 10300) PubDate.second>=20.5 84 23 N (0.72619048 0.27380952)
## 20600) PubDate.month< 10.5 62 12 N (0.80645161 0.19354839)
## 41200) PubDate.month>=9.5 28 1 N (0.96428571 0.03571429) *
## 41201) PubDate.month< 9.5 34 11 N (0.67647059 0.32352941)
## 82402) A.num.chars< 172 27 6 N (0.77777778 0.22222222) *
## 82403) A.num.chars>=172 7 2 Y (0.28571429 0.71428571) *
## 20601) PubDate.month>=10.5 22 11 N (0.50000000 0.50000000)
## 41202) PubDate.date< 10.5 9 2 N (0.77777778 0.22222222) *
## 41203) PubDate.date>=10.5 13 4 Y (0.30769231 0.69230769) *
## 10301) PubDate.second< 20.5 42 21 N (0.50000000 0.50000000)
## 20602) A.num.chars>=128.5 25 8 N (0.68000000 0.32000000)
## 41204) PubDate.month< 10.5 17 3 N (0.82352941 0.17647059) *
## 41205) PubDate.month>=10.5 8 3 Y (0.37500000 0.62500000) *
## 20603) A.num.chars< 128.5 17 4 Y (0.23529412 0.76470588) *
## 5151) H.num.chars< 21.5 10 2 Y (0.20000000 0.80000000) *
## 161) SubsectionName.fctrThe Public Editor>=0.5 11 2 Y (0.18181818 0.81818182) *
## 81) PubDate.wkday>=5.5 17 4 Y (0.23529412 0.76470588) *
## 41) NewsDesk.fctrStyles>=0.5 78 35 Y (0.44871795 0.55128205)
## 82) SectionName.fctr>=0.5 13 0 N (1.00000000 0.00000000) *
## 83) SectionName.fctr< 0.5 65 22 Y (0.33846154 0.66153846)
## 166) A.num.chars< 174.5 40 18 Y (0.45000000 0.55000000)
## 332) PubDate.date>=17.5 17 6 N (0.64705882 0.35294118) *
## 333) PubDate.date< 17.5 23 7 Y (0.30434783 0.69565217)
## 666) PubDate.minute>=38 7 2 N (0.71428571 0.28571429) *
## 667) PubDate.minute< 38 16 2 Y (0.12500000 0.87500000) *
## 167) A.num.chars>=174.5 25 4 Y (0.16000000 0.84000000) *
## 21) PubDate.hour>=21.5 65 22 Y (0.33846154 0.66153846)
## 42) A.num.chars>=92 25 8 N (0.68000000 0.32000000)
## 84) PubDate.date>=11 18 4 N (0.77777778 0.22222222) *
## 85) PubDate.date< 11 7 3 Y (0.42857143 0.57142857) *
## 43) A.num.chars< 92 40 5 Y (0.12500000 0.87500000) *
## 11) SectionName.fctrHealth>=0.5 71 6 Y (0.08450704 0.91549296) *
## 3) NewsDesk.fctrOpEd>=0.5 344 75 Y (0.21802326 0.78197674)
## 6) A.num.chars>=169.5 8 1 N (0.87500000 0.12500000) *
## 7) A.num.chars< 169.5 336 68 Y (0.20238095 0.79761905)
## 14) WordCount< 107 34 14 Y (0.41176471 0.58823529)
## 28) PubDate.wkday>=2.5 23 10 N (0.56521739 0.43478261)
## 56) PubDate.hour>=15.5 11 1 N (0.90909091 0.09090909) *
## 57) PubDate.hour< 15.5 12 3 Y (0.25000000 0.75000000) *
## 29) PubDate.wkday< 2.5 11 1 Y (0.09090909 0.90909091) *
## 15) WordCount>=107 302 54 Y (0.17880795 0.82119205)
## 30) A.num.chars>=38.5 254 52 Y (0.20472441 0.79527559)
## 60) PubDate.wkday>=3.5 105 28 Y (0.26666667 0.73333333)
## 120) WordCount< 483 48 18 Y (0.37500000 0.62500000)
## 240) A.num.chars>=56 41 18 Y (0.43902439 0.56097561)
## 480) A.num.chars< 70.5 10 2 N (0.80000000 0.20000000) *
## 481) A.num.chars>=70.5 31 10 Y (0.32258065 0.67741935)
## 962) H.num.chars< 39 12 5 N (0.58333333 0.41666667) *
## 963) H.num.chars>=39 19 3 Y (0.15789474 0.84210526) *
## 241) A.num.chars< 56 7 0 Y (0.00000000 1.00000000) *
## 121) WordCount>=483 57 10 Y (0.17543860 0.82456140) *
## 61) PubDate.wkday< 3.5 149 24 Y (0.16107383 0.83892617) *
## 31) A.num.chars< 38.5 48 2 Y (0.04166667 0.95833333) *
## [1] " calling mypredict_mdl for fit:"
## Reference
## Prediction N Y
## N 0 0
## Y 3726 749
## Popular.fctr Popular.fctr.predict.Conditional.X.cp.0.rpart.N
## 1 N 0
## 2 Y 0
## Popular.fctr.predict.Conditional.X.cp.0.rpart.Y
## 1 3726
## 2 749
## Reference
## Prediction N Y
## N 3269 78
## Y 457 671
## Popular.fctr Popular.fctr.predict.Conditional.X.cp.0.rpart.N
## 1 N 3269
## 2 Y 78
## Popular.fctr.predict.Conditional.X.cp.0.rpart.Y
## 1 457
## 2 671
## Reference
## Prediction N Y
## N 3413 105
## Y 313 644
## Popular.fctr Popular.fctr.predict.Conditional.X.cp.0.rpart.N
## 1 N 3413
## 2 Y 105
## Popular.fctr.predict.Conditional.X.cp.0.rpart.Y
## 1 313
## 2 644
## Reference
## Prediction N Y
## N 3565 151
## Y 161 598
## Popular.fctr Popular.fctr.predict.Conditional.X.cp.0.rpart.N
## 1 N 3565
## 2 Y 151
## Popular.fctr.predict.Conditional.X.cp.0.rpart.Y
## 1 161
## 2 598
## Reference
## Prediction N Y
## N 3586 162
## Y 140 587
## Popular.fctr Popular.fctr.predict.Conditional.X.cp.0.rpart.N
## 1 N 3586
## 2 Y 162
## Popular.fctr.predict.Conditional.X.cp.0.rpart.Y
## 1 140
## 2 587
## Reference
## Prediction N Y
## N 3597 170
## Y 129 579
## Popular.fctr Popular.fctr.predict.Conditional.X.cp.0.rpart.N
## 1 N 3597
## 2 Y 170
## Popular.fctr.predict.Conditional.X.cp.0.rpart.Y
## 1 129
## 2 579
## Reference
## Prediction N Y
## N 3609 186
## Y 117 563
## Popular.fctr Popular.fctr.predict.Conditional.X.cp.0.rpart.N
## 1 N 3609
## 2 Y 186
## Popular.fctr.predict.Conditional.X.cp.0.rpart.Y
## 1 117
## 2 563
## Reference
## Prediction N Y
## N 3637 238
## Y 89 511
## Popular.fctr Popular.fctr.predict.Conditional.X.cp.0.rpart.N
## 1 N 3637
## 2 Y 238
## Popular.fctr.predict.Conditional.X.cp.0.rpart.Y
## 1 89
## 2 511
## Reference
## Prediction N Y
## N 3665 320
## Y 61 429
## Popular.fctr Popular.fctr.predict.Conditional.X.cp.0.rpart.N
## 1 N 3665
## 2 Y 320
## Popular.fctr.predict.Conditional.X.cp.0.rpart.Y
## 1 61
## 2 429
## Reference
## Prediction N Y
## N 3716 595
## Y 10 154
## Popular.fctr Popular.fctr.predict.Conditional.X.cp.0.rpart.N
## 1 N 3716
## 2 Y 595
## Popular.fctr.predict.Conditional.X.cp.0.rpart.Y
## 1 10
## 2 154
## Reference
## Prediction N Y
## N 3726 742
## Y 0 7
## Popular.fctr Popular.fctr.predict.Conditional.X.cp.0.rpart.N
## 1 N 3726
## 2 Y 742
## Popular.fctr.predict.Conditional.X.cp.0.rpart.Y
## 1 0
## 2 7
## threshold f.score
## 1 0.0 0.28675345
## 2 0.1 0.71497070
## 3 0.2 0.75498242
## 4 0.3 0.79310345
## 5 0.4 0.79539295
## 6 0.5 0.79478380
## 7 0.6 0.78796361
## 8 0.7 0.75759822
## 9 0.8 0.69249395
## 10 0.9 0.33734940
## 11 1.0 0.01851852
## [1] "Classifier Probability Threshold: 0.4000 to maximize f.score.fit"
## Popular.fctr Popular.fctr.predict.Conditional.X.cp.0.rpart.N
## 1 N 3586
## 2 Y 162
## Popular.fctr.predict.Conditional.X.cp.0.rpart.Y
## 1 140
## 2 587
## Reference
## Prediction N Y
## N 3586 162
## Y 140 587
## Popular.fctr Popular.fctr.predict.Conditional.X.cp.0.rpart.N
## 1 N 3586
## 2 Y 162
## Popular.fctr.predict.Conditional.X.cp.0.rpart.Y
## 1 140
## 2 587
## [1] "in mypredict_mdl: calling confusionMatrix..."
## [1] " unique(df[, rsp_var_out]: "
## [1] Y N
## Levels: N Y
## [1] " unique(df[, rsp_var ]: "
## [1] Y N
## Levels: N Y
## Prediction
## Reference N Y
## N 3586 140
## Y 162 587
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 9.325140e-01 7.549970e-01 9.247659e-01 9.396915e-01 8.326257e-01
## AccuracyPValue McnemarPValue
## 2.322542e-88 2.268880e-01
## [1] " calling mypredict_mdl for OOB:"
## Reference
## Prediction N Y
## N 0 0
## Y 1713 344
## Popular.fctr Popular.fctr.predict.Conditional.X.cp.0.rpart.N
## 1 N 0
## 2 Y 0
## Popular.fctr.predict.Conditional.X.cp.0.rpart.Y
## 1 1713
## 2 344
## Reference
## Prediction N Y
## N 1458 77
## Y 255 267
## Popular.fctr Popular.fctr.predict.Conditional.X.cp.0.rpart.N
## 1 N 1458
## 2 Y 77
## Popular.fctr.predict.Conditional.X.cp.0.rpart.Y
## 1 255
## 2 267
## Reference
## Prediction N Y
## N 1520 93
## Y 193 251
## Popular.fctr Popular.fctr.predict.Conditional.X.cp.0.rpart.N
## 1 N 1520
## 2 Y 93
## Popular.fctr.predict.Conditional.X.cp.0.rpart.Y
## 1 193
## 2 251
## Reference
## Prediction N Y
## N 1583 116
## Y 130 228
## Popular.fctr Popular.fctr.predict.Conditional.X.cp.0.rpart.N
## 1 N 1583
## 2 Y 116
## Popular.fctr.predict.Conditional.X.cp.0.rpart.Y
## 1 130
## 2 228
## Reference
## Prediction N Y
## N 1594 119
## Y 119 225
## Popular.fctr Popular.fctr.predict.Conditional.X.cp.0.rpart.N
## 1 N 1594
## 2 Y 119
## Popular.fctr.predict.Conditional.X.cp.0.rpart.Y
## 1 119
## 2 225
## Reference
## Prediction N Y
## N 1597 124
## Y 116 220
## Popular.fctr Popular.fctr.predict.Conditional.X.cp.0.rpart.N
## 1 N 1597
## 2 Y 124
## Popular.fctr.predict.Conditional.X.cp.0.rpart.Y
## 1 116
## 2 220
## Reference
## Prediction N Y
## N 1608 125
## Y 105 219
## Popular.fctr Popular.fctr.predict.Conditional.X.cp.0.rpart.N
## 1 N 1608
## 2 Y 125
## Popular.fctr.predict.Conditional.X.cp.0.rpart.Y
## 1 105
## 2 219
## Reference
## Prediction N Y
## N 1629 137
## Y 84 207
## Popular.fctr Popular.fctr.predict.Conditional.X.cp.0.rpart.N
## 1 N 1629
## 2 Y 137
## Popular.fctr.predict.Conditional.X.cp.0.rpart.Y
## 1 84
## 2 207
## Reference
## Prediction N Y
## N 1666 153
## Y 47 191
## Popular.fctr Popular.fctr.predict.Conditional.X.cp.0.rpart.N
## 1 N 1666
## 2 Y 153
## Popular.fctr.predict.Conditional.X.cp.0.rpart.Y
## 1 47
## 2 191
## Reference
## Prediction N Y
## N 1703 277
## Y 10 67
## Popular.fctr Popular.fctr.predict.Conditional.X.cp.0.rpart.N
## 1 N 1703
## 2 Y 277
## Popular.fctr.predict.Conditional.X.cp.0.rpart.Y
## 1 10
## 2 67
## Reference
## Prediction N Y
## N 1712 340
## Y 1 4
## Popular.fctr Popular.fctr.predict.Conditional.X.cp.0.rpart.N
## 1 N 1712
## 2 Y 340
## Popular.fctr.predict.Conditional.X.cp.0.rpart.Y
## 1 1
## 2 4
## threshold f.score
## 1 0.0 0.28654727
## 2 0.1 0.61662818
## 3 0.2 0.63705584
## 4 0.3 0.64957265
## 5 0.4 0.65406977
## 6 0.5 0.64705882
## 7 0.6 0.65568862
## 8 0.7 0.65196850
## 9 0.8 0.65635739
## 10 0.9 0.31828979
## 11 1.0 0.02292264## [1] "Classifier Probability Threshold: 0.8000 to maximize f.score.OOB"
## Popular.fctr Popular.fctr.predict.Conditional.X.cp.0.rpart.N
## 1 N 1666
## 2 Y 153
## Popular.fctr.predict.Conditional.X.cp.0.rpart.Y
## 1 47
## 2 191
## Reference
## Prediction N Y
## N 1666 153
## Y 47 191
## Popular.fctr Popular.fctr.predict.Conditional.X.cp.0.rpart.N
## 1 N 1666
## 2 Y 153
## Popular.fctr.predict.Conditional.X.cp.0.rpart.Y
## 1 47
## 2 191
## [1] "in mypredict_mdl: calling confusionMatrix..."
## [1] " unique(df[, rsp_var_out]: "
## [1] N Y
## Levels: N Y
## [1] " unique(df[, rsp_var ]: "
## [1] N Y
## Levels: N Y
## Prediction
## Reference N Y
## N 1666 47
## Y 153 191
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 9.027710e-01 6.019082e-01 8.891448e-01 9.152364e-01 8.327662e-01
## AccuracyPValue McnemarPValue
## 6.489691e-20 1.131031e-13
## model_id model_method
## 1 Conditional.X.cp.0.rpart rpart
## feats
## 1 WordCount, PubDate.hour, PubDate.month, PubDate.date, PubDate.second, PubDate.minute, PubDate.wkday, S.year, A.year, S.compani, A.compani, S.time, A.time, S.will, A.will, S.york, A.york, S.new, A.new, H.new, S.week, A.week, SectionName.fctr, NewsDesk.fctr, SubsectionName.fctr, H.num.chars, H.num.chars.log, A.num.chars, S.num.chars, H.num.words, H.num.words.unq, H.num.words.log, S.num.chars.log, A.num.chars.log, A.num.words, S.num.words, H.num.words.unq.log, A.num.words.unq, S.num.words.unq, A.num.words.log, S.num.words.log, A.num.words.unq.log, S.num.words.unq.log
## max.nTuningRuns min.elapsedtime.everything min.elapsedtime.final
## 1 0 1.525 0.897
## max.auc.fit opt.prob.threshold.fit max.f.score.fit max.Accuracy.fit
## 1 0.9468583 0.4 0.795393 0.932514
## max.AccuracyLower.fit max.AccuracyUpper.fit max.Kappa.fit max.auc.OOB
## 1 0.9247659 0.9396915 0.754997 0.8910062
## opt.prob.threshold.OOB max.f.score.OOB max.Accuracy.OOB
## 1 0.8 0.6563574 0.902771
## max.AccuracyLower.OOB max.AccuracyUpper.OOB max.Kappa.OOB
## 1 0.8891448 0.9152364 0.6019082
## [1] "fitting model: Conditional.X.rf"
## [1] " indep_vars: WordCount, PubDate.hour, PubDate.month, .rnorm, PubDate.date, PubDate.second, PubDate.minute, PubDate.wkday, S.year, A.year, S.compani, A.compani, S.time, A.time, S.will, A.will, S.york, A.york, S.new, A.new, H.new, S.week, A.week, SectionName.fctr, NewsDesk.fctr, SubsectionName.fctr, H.num.chars, H.num.chars.log, A.num.chars, S.num.chars, H.num.words, H.num.words.unq, H.num.words.log, S.num.chars.log, A.num.chars.log, A.num.words, S.num.words, H.num.words.unq.log, A.num.words.unq, S.num.words.unq, A.num.words.log, S.num.words.log, A.num.words.unq.log, S.num.words.unq.log"## Loading required package: randomForest
## randomForest 4.6-10
## Type rfNews() to see new features/changes/bug fixes.
## + : mtry= 2
## - : mtry= 2
## + : mtry=39
## - : mtry=39
## + : mtry=76
## - : mtry=76
## Aggregating results
## Selecting tuning parameters
## Fitting mtry = 39 on full training set
## Length Class Mode
## call 4 -none- call
## type 1 -none- character
## predicted 4475 factor numeric
## err.rate 1500 -none- numeric
## confusion 6 -none- numeric
## votes 8950 matrix numeric
## oob.times 4475 -none- numeric
## classes 2 -none- character
## importance 76 -none- numeric
## importanceSD 0 -none- NULL
## localImportance 0 -none- NULL
## proximity 0 -none- NULL
## ntree 1 -none- numeric
## mtry 1 -none- numeric
## forest 14 -none- list
## y 4475 factor numeric
## test 0 -none- NULL
## inbag 0 -none- NULL
## xNames 76 -none- character
## problemType 1 -none- character
## tuneValue 1 data.frame list
## obsLevels 2 -none- character
## [1] " calling mypredict_mdl for fit:"
## Reference
## Prediction N Y
## N 0 0
## Y 3726 749
## Popular.fctr Popular.fctr.predict.Conditional.X.rf.N
## 1 N 0
## 2 Y 0
## Popular.fctr.predict.Conditional.X.rf.Y
## 1 3726
## 2 749
## Reference
## Prediction N Y
## N 3341 0
## Y 385 749
## Popular.fctr Popular.fctr.predict.Conditional.X.rf.N
## 1 N 3341
## 2 Y 0
## Popular.fctr.predict.Conditional.X.rf.Y
## 1 385
## 2 749
## Reference
## Prediction N Y
## N 3607 0
## Y 119 749
## Popular.fctr Popular.fctr.predict.Conditional.X.rf.N
## 1 N 3607
## 2 Y 0
## Popular.fctr.predict.Conditional.X.rf.Y
## 1 119
## 2 749
## Reference
## Prediction N Y
## N 3690 0
## Y 36 749
## Popular.fctr Popular.fctr.predict.Conditional.X.rf.N
## 1 N 3690
## 2 Y 0
## Popular.fctr.predict.Conditional.X.rf.Y
## 1 36
## 2 749
## Reference
## Prediction N Y
## N 3726 0
## Y 0 749
## Popular.fctr Popular.fctr.predict.Conditional.X.rf.N
## 1 N 3726
## 2 Y 0
## Popular.fctr.predict.Conditional.X.rf.Y
## 1 0
## 2 749
## Reference
## Prediction N Y
## N 3726 0
## Y 0 749
## Popular.fctr Popular.fctr.predict.Conditional.X.rf.N
## 1 N 3726
## 2 Y 0
## Popular.fctr.predict.Conditional.X.rf.Y
## 1 0
## 2 749
## Reference
## Prediction N Y
## N 3726 2
## Y 0 747
## Popular.fctr Popular.fctr.predict.Conditional.X.rf.N
## 1 N 3726
## 2 Y 2
## Popular.fctr.predict.Conditional.X.rf.Y
## 1 0
## 2 747
## Reference
## Prediction N Y
## N 3726 94
## Y 0 655
## Popular.fctr Popular.fctr.predict.Conditional.X.rf.N
## 1 N 3726
## 2 Y 94
## Popular.fctr.predict.Conditional.X.rf.Y
## 1 0
## 2 655
## Reference
## Prediction N Y
## N 3726 247
## Y 0 502
## Popular.fctr Popular.fctr.predict.Conditional.X.rf.N
## 1 N 3726
## 2 Y 247
## Popular.fctr.predict.Conditional.X.rf.Y
## 1 0
## 2 502
## Reference
## Prediction N Y
## N 3726 409
## Y 0 340
## Popular.fctr Popular.fctr.predict.Conditional.X.rf.N
## 1 N 3726
## 2 Y 409
## Popular.fctr.predict.Conditional.X.rf.Y
## 1 0
## 2 340
## Reference
## Prediction N Y
## N 3726 747
## Y 0 2
## Popular.fctr Popular.fctr.predict.Conditional.X.rf.N
## 1 N 3726
## 2 Y 747
## Popular.fctr.predict.Conditional.X.rf.Y
## 1 0
## 2 2
## threshold f.score
## 1 0.0 0.286753446
## 2 0.1 0.795539033
## 3 0.2 0.926406926
## 4 0.3 0.976531943
## 5 0.4 1.000000000
## 6 0.5 1.000000000
## 7 0.6 0.998663102
## 8 0.7 0.933048433
## 9 0.8 0.802557954
## 10 0.9 0.624426079
## 11 1.0 0.005326232
## [1] "Classifier Probability Threshold: 0.5000 to maximize f.score.fit"
## Popular.fctr Popular.fctr.predict.Conditional.X.rf.N
## 1 N 3726
## 2 Y NA
## Popular.fctr.predict.Conditional.X.rf.Y
## 1 NA
## 2 749
## Reference
## Prediction N Y
## N 3726 0
## Y 0 749
## Popular.fctr Popular.fctr.predict.Conditional.X.rf.N
## 1 N 3726
## 2 Y 0
## Popular.fctr.predict.Conditional.X.rf.Y
## 1 0
## 2 749
## [1] "in mypredict_mdl: calling confusionMatrix..."
## [1] " unique(df[, rsp_var_out]: "
## [1] Y N
## Levels: N Y
## [1] " unique(df[, rsp_var ]: "
## [1] Y N
## Levels: N Y
## Prediction
## Reference N Y
## N 3726 0
## Y 0 749
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 1.0000000 1.0000000 0.9991760 1.0000000 0.8326257
## AccuracyPValue McnemarPValue
## 0.0000000 NaN
## [1] " calling mypredict_mdl for OOB:"
## Reference
## Prediction N Y
## N 0 0
## Y 1713 344
## Popular.fctr Popular.fctr.predict.Conditional.X.rf.N
## 1 N 0
## 2 Y 0
## Popular.fctr.predict.Conditional.X.rf.Y
## 1 1713
## 2 344
## Reference
## Prediction N Y
## N 1271 24
## Y 442 320
## Popular.fctr Popular.fctr.predict.Conditional.X.rf.N
## 1 N 1271
## 2 Y 24
## Popular.fctr.predict.Conditional.X.rf.Y
## 1 442
## 2 320
## Reference
## Prediction N Y
## N 1449 50
## Y 264 294
## Popular.fctr Popular.fctr.predict.Conditional.X.rf.N
## 1 N 1449
## 2 Y 50
## Popular.fctr.predict.Conditional.X.rf.Y
## 1 264
## 2 294
## Reference
## Prediction N Y
## N 1564 72
## Y 149 272
## Popular.fctr Popular.fctr.predict.Conditional.X.rf.N
## 1 N 1564
## 2 Y 72
## Popular.fctr.predict.Conditional.X.rf.Y
## 1 149
## 2 272
## Reference
## Prediction N Y
## N 1621 99
## Y 92 245
## Popular.fctr Popular.fctr.predict.Conditional.X.rf.N
## 1 N 1621
## 2 Y 99
## Popular.fctr.predict.Conditional.X.rf.Y
## 1 92
## 2 245
## Reference
## Prediction N Y
## N 1653 120
## Y 60 224
## Popular.fctr Popular.fctr.predict.Conditional.X.rf.N
## 1 N 1653
## 2 Y 120
## Popular.fctr.predict.Conditional.X.rf.Y
## 1 60
## 2 224
## Reference
## Prediction N Y
## N 1667 149
## Y 46 195
## Popular.fctr Popular.fctr.predict.Conditional.X.rf.N
## 1 N 1667
## 2 Y 149
## Popular.fctr.predict.Conditional.X.rf.Y
## 1 46
## 2 195
## Reference
## Prediction N Y
## N 1682 181
## Y 31 163
## Popular.fctr Popular.fctr.predict.Conditional.X.rf.N
## 1 N 1682
## 2 Y 181
## Popular.fctr.predict.Conditional.X.rf.Y
## 1 31
## 2 163
## Reference
## Prediction N Y
## N 1697 225
## Y 16 119
## Popular.fctr Popular.fctr.predict.Conditional.X.rf.N
## 1 N 1697
## 2 Y 225
## Popular.fctr.predict.Conditional.X.rf.Y
## 1 16
## 2 119
## Reference
## Prediction N Y
## N 1707 290
## Y 6 54
## Popular.fctr Popular.fctr.predict.Conditional.X.rf.N
## 1 N 1707
## 2 Y 290
## Popular.fctr.predict.Conditional.X.rf.Y
## 1 6
## 2 54
## Reference
## Prediction N Y
## N 1713 344
## Y 0 0
## Popular.fctr Popular.fctr.predict.Conditional.X.rf.N
## 1 N 1713
## 2 Y 344
## Popular.fctr.predict.Conditional.X.rf.Y
## 1 0
## 2 0
## threshold f.score
## 1 0.0 0.2865473
## 2 0.1 0.5786618
## 3 0.2 0.6518847
## 4 0.3 0.7111111
## 5 0.4 0.7195301
## 6 0.5 0.7133758
## 7 0.6 0.6666667
## 8 0.7 0.6059480
## 9 0.8 0.4968685
## 10 0.9 0.2673267
## 11 1.0 0.0000000
## [1] "Classifier Probability Threshold: 0.4000 to maximize f.score.OOB"
## Popular.fctr Popular.fctr.predict.Conditional.X.rf.N
## 1 N 1621
## 2 Y 99
## Popular.fctr.predict.Conditional.X.rf.Y
## 1 92
## 2 245
## Reference
## Prediction N Y
## N 1621 99
## Y 92 245
## Popular.fctr Popular.fctr.predict.Conditional.X.rf.N
## 1 N 1621
## 2 Y 99
## Popular.fctr.predict.Conditional.X.rf.Y
## 1 92
## 2 245
## [1] "in mypredict_mdl: calling confusionMatrix..."
## [1] " unique(df[, rsp_var_out]: "
## [1] N Y
## Levels: N Y
## [1] " unique(df[, rsp_var ]: "
## [1] N Y
## Levels: N Y
## Prediction
## Reference N Y
## N 1621 92
## Y 99 245
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 9.071463e-01 6.639007e-01 8.937741e-01 9.193445e-01 8.327662e-01
## AccuracyPValue McnemarPValue
## 1.834204e-22 6.641833e-01
## model_id model_method
## 1 Conditional.X.rf rf
## feats
## 1 WordCount, PubDate.hour, PubDate.month, .rnorm, PubDate.date, PubDate.second, PubDate.minute, PubDate.wkday, S.year, A.year, S.compani, A.compani, S.time, A.time, S.will, A.will, S.york, A.york, S.new, A.new, H.new, S.week, A.week, SectionName.fctr, NewsDesk.fctr, SubsectionName.fctr, H.num.chars, H.num.chars.log, A.num.chars, S.num.chars, H.num.words, H.num.words.unq, H.num.words.log, S.num.chars.log, A.num.chars.log, A.num.words, S.num.words, H.num.words.unq.log, A.num.words.unq, S.num.words.unq, A.num.words.log, S.num.words.log, A.num.words.unq.log, S.num.words.unq.log
## max.nTuningRuns min.elapsedtime.everything min.elapsedtime.final
## 1 3 105.229 25.835
## max.auc.fit opt.prob.threshold.fit max.f.score.fit max.Accuracy.fit
## 1 1 0.5 1 0.9086034
## max.AccuracyLower.fit max.AccuracyUpper.fit max.Kappa.fit max.auc.OOB
## 1 0.999176 1 0.6478558 0.9287358
## opt.prob.threshold.OOB max.f.score.OOB max.Accuracy.OOB
## 1 0.4 0.7195301 0.9071463
## max.AccuracyLower.OOB max.AccuracyUpper.OOB max.Kappa.OOB
## 1 0.8937741 0.9193445 0.6639007
## [1] "fitting model: Conditional.X.no.rnorm.rf"
## [1] " indep_vars: WordCount, PubDate.hour, PubDate.month, PubDate.date, PubDate.second, PubDate.minute, PubDate.wkday, S.year, A.year, S.compani, A.compani, S.time, A.time, S.will, A.will, S.york, A.york, S.new, A.new, H.new, S.week, A.week, SectionName.fctr, NewsDesk.fctr, SubsectionName.fctr, H.num.chars, H.num.chars.log, A.num.chars, S.num.chars, H.num.words, H.num.words.unq, H.num.words.log, S.num.chars.log, A.num.chars.log, A.num.words, S.num.words, H.num.words.unq.log, A.num.words.unq, S.num.words.unq, A.num.words.log, S.num.words.log, A.num.words.unq.log, S.num.words.unq.log"
## + : mtry= 2
## - : mtry= 2
## + : mtry=38
## - : mtry=38
## + : mtry=75
## - : mtry=75
## Aggregating results
## Selecting tuning parameters
## Fitting mtry = 38 on full training set
## Length Class Mode
## call 4 -none- call
## type 1 -none- character
## predicted 4475 factor numeric
## err.rate 1500 -none- numeric
## confusion 6 -none- numeric
## votes 8950 matrix numeric
## oob.times 4475 -none- numeric
## classes 2 -none- character
## importance 75 -none- numeric
## importanceSD 0 -none- NULL
## localImportance 0 -none- NULL
## proximity 0 -none- NULL
## ntree 1 -none- numeric
## mtry 1 -none- numeric
## forest 14 -none- list
## y 4475 factor numeric
## test 0 -none- NULL
## inbag 0 -none- NULL
## xNames 75 -none- character
## problemType 1 -none- character
## tuneValue 1 data.frame list
## obsLevels 2 -none- character
## [1] " calling mypredict_mdl for fit:"
## Reference
## Prediction N Y
## N 0 0
## Y 3726 749
## Popular.fctr Popular.fctr.predict.Conditional.X.no.rnorm.rf.N
## 1 N 0
## 2 Y 0
## Popular.fctr.predict.Conditional.X.no.rnorm.rf.Y
## 1 3726
## 2 749
## Reference
## Prediction N Y
## N 3346 0
## Y 380 749
## Popular.fctr Popular.fctr.predict.Conditional.X.no.rnorm.rf.N
## 1 N 3346
## 2 Y 0
## Popular.fctr.predict.Conditional.X.no.rnorm.rf.Y
## 1 380
## 2 749
## Reference
## Prediction N Y
## N 3605 0
## Y 121 749
## Popular.fctr Popular.fctr.predict.Conditional.X.no.rnorm.rf.N
## 1 N 3605
## 2 Y 0
## Popular.fctr.predict.Conditional.X.no.rnorm.rf.Y
## 1 121
## 2 749
## Reference
## Prediction N Y
## N 3689 0
## Y 37 749
## Popular.fctr Popular.fctr.predict.Conditional.X.no.rnorm.rf.N
## 1 N 3689
## 2 Y 0
## Popular.fctr.predict.Conditional.X.no.rnorm.rf.Y
## 1 37
## 2 749
## Reference
## Prediction N Y
## N 3725 0
## Y 1 749
## Popular.fctr Popular.fctr.predict.Conditional.X.no.rnorm.rf.N
## 1 N 3725
## 2 Y 0
## Popular.fctr.predict.Conditional.X.no.rnorm.rf.Y
## 1 1
## 2 749
## Reference
## Prediction N Y
## N 3726 0
## Y 0 749
## Popular.fctr Popular.fctr.predict.Conditional.X.no.rnorm.rf.N
## 1 N 3726
## 2 Y 0
## Popular.fctr.predict.Conditional.X.no.rnorm.rf.Y
## 1 0
## 2 749
## Reference
## Prediction N Y
## N 3726 0
## Y 0 749
## Popular.fctr Popular.fctr.predict.Conditional.X.no.rnorm.rf.N
## 1 N 3726
## 2 Y 0
## Popular.fctr.predict.Conditional.X.no.rnorm.rf.Y
## 1 0
## 2 749
## Reference
## Prediction N Y
## N 3726 96
## Y 0 653
## Popular.fctr Popular.fctr.predict.Conditional.X.no.rnorm.rf.N
## 1 N 3726
## 2 Y 96
## Popular.fctr.predict.Conditional.X.no.rnorm.rf.Y
## 1 0
## 2 653
## Reference
## Prediction N Y
## N 3726 252
## Y 0 497
## Popular.fctr Popular.fctr.predict.Conditional.X.no.rnorm.rf.N
## 1 N 3726
## 2 Y 252
## Popular.fctr.predict.Conditional.X.no.rnorm.rf.Y
## 1 0
## 2 497
## Reference
## Prediction N Y
## N 3726 408
## Y 0 341
## Popular.fctr Popular.fctr.predict.Conditional.X.no.rnorm.rf.N
## 1 N 3726
## 2 Y 408
## Popular.fctr.predict.Conditional.X.no.rnorm.rf.Y
## 1 0
## 2 341
## Reference
## Prediction N Y
## N 3726 748
## Y 0 1
## Popular.fctr Popular.fctr.predict.Conditional.X.no.rnorm.rf.N
## 1 N 3726
## 2 Y 748
## Popular.fctr.predict.Conditional.X.no.rnorm.rf.Y
## 1 0
## 2 1
## threshold f.score
## 1 0.0 0.286753446
## 2 0.1 0.797657082
## 3 0.2 0.925262508
## 4 0.3 0.975895765
## 5 0.4 0.999332889
## 6 0.5 1.000000000
## 7 0.6 1.000000000
## 8 0.7 0.931526391
## 9 0.8 0.797752809
## 10 0.9 0.625688073
## 11 1.0 0.002666667
## [1] "Classifier Probability Threshold: 0.6000 to maximize f.score.fit"
## Popular.fctr Popular.fctr.predict.Conditional.X.no.rnorm.rf.N
## 1 N 3726
## 2 Y NA
## Popular.fctr.predict.Conditional.X.no.rnorm.rf.Y
## 1 NA
## 2 749
## Reference
## Prediction N Y
## N 3726 0
## Y 0 749
## Popular.fctr Popular.fctr.predict.Conditional.X.no.rnorm.rf.N
## 1 N 3726
## 2 Y 0
## Popular.fctr.predict.Conditional.X.no.rnorm.rf.Y
## 1 0
## 2 749
## [1] "in mypredict_mdl: calling confusionMatrix..."
## [1] " unique(df[, rsp_var_out]: "
## [1] Y N
## Levels: N Y
## [1] " unique(df[, rsp_var ]: "
## [1] Y N
## Levels: N Y
## Prediction
## Reference N Y
## N 3726 0
## Y 0 749
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 1.0000000 1.0000000 0.9991760 1.0000000 0.8326257
## AccuracyPValue McnemarPValue
## 0.0000000 NaN
## [1] " calling mypredict_mdl for OOB:"
## Reference
## Prediction N Y
## N 0 0
## Y 1713 344
## Popular.fctr Popular.fctr.predict.Conditional.X.no.rnorm.rf.N
## 1 N 0
## 2 Y 0
## Popular.fctr.predict.Conditional.X.no.rnorm.rf.Y
## 1 1713
## 2 344
## Reference
## Prediction N Y
## N 1277 25
## Y 436 319
## Popular.fctr Popular.fctr.predict.Conditional.X.no.rnorm.rf.N
## 1 N 1277
## 2 Y 25
## Popular.fctr.predict.Conditional.X.no.rnorm.rf.Y
## 1 436
## 2 319
## Reference
## Prediction N Y
## N 1454 50
## Y 259 294
## Popular.fctr Popular.fctr.predict.Conditional.X.no.rnorm.rf.N
## 1 N 1454
## 2 Y 50
## Popular.fctr.predict.Conditional.X.no.rnorm.rf.Y
## 1 259
## 2 294
## Reference
## Prediction N Y
## N 1569 71
## Y 144 273
## Popular.fctr Popular.fctr.predict.Conditional.X.no.rnorm.rf.N
## 1 N 1569
## 2 Y 71
## Popular.fctr.predict.Conditional.X.no.rnorm.rf.Y
## 1 144
## 2 273
## Reference
## Prediction N Y
## N 1621 98
## Y 92 246
## Popular.fctr Popular.fctr.predict.Conditional.X.no.rnorm.rf.N
## 1 N 1621
## 2 Y 98
## Popular.fctr.predict.Conditional.X.no.rnorm.rf.Y
## 1 92
## 2 246
## Reference
## Prediction N Y
## N 1653 121
## Y 60 223
## Popular.fctr Popular.fctr.predict.Conditional.X.no.rnorm.rf.N
## 1 N 1653
## 2 Y 121
## Popular.fctr.predict.Conditional.X.no.rnorm.rf.Y
## 1 60
## 2 223
## Reference
## Prediction N Y
## N 1668 150
## Y 45 194
## Popular.fctr Popular.fctr.predict.Conditional.X.no.rnorm.rf.N
## 1 N 1668
## 2 Y 150
## Popular.fctr.predict.Conditional.X.no.rnorm.rf.Y
## 1 45
## 2 194
## Reference
## Prediction N Y
## N 1679 180
## Y 34 164
## Popular.fctr Popular.fctr.predict.Conditional.X.no.rnorm.rf.N
## 1 N 1679
## 2 Y 180
## Popular.fctr.predict.Conditional.X.no.rnorm.rf.Y
## 1 34
## 2 164
## Reference
## Prediction N Y
## N 1698 224
## Y 15 120
## Popular.fctr Popular.fctr.predict.Conditional.X.no.rnorm.rf.N
## 1 N 1698
## 2 Y 224
## Popular.fctr.predict.Conditional.X.no.rnorm.rf.Y
## 1 15
## 2 120
## Reference
## Prediction N Y
## N 1707 282
## Y 6 62
## Popular.fctr Popular.fctr.predict.Conditional.X.no.rnorm.rf.N
## 1 N 1707
## 2 Y 282
## Popular.fctr.predict.Conditional.X.no.rnorm.rf.Y
## 1 6
## 2 62
## Reference
## Prediction N Y
## N 1713 344
## Y 0 0
## Popular.fctr Popular.fctr.predict.Conditional.X.no.rnorm.rf.N
## 1 N 1713
## 2 Y 344
## Popular.fctr.predict.Conditional.X.no.rnorm.rf.Y
## 1 0
## 2 0
## threshold f.score
## 1 0.0 0.2865473
## 2 0.1 0.5805278
## 3 0.2 0.6555184
## 4 0.3 0.7174770
## 5 0.4 0.7214076
## 6 0.5 0.7113238
## 7 0.6 0.6655232
## 8 0.7 0.6051661
## 9 0.8 0.5010438
## 10 0.9 0.3009709
## 11 1.0 0.0000000
## [1] "Classifier Probability Threshold: 0.4000 to maximize f.score.OOB"
## Popular.fctr Popular.fctr.predict.Conditional.X.no.rnorm.rf.N
## 1 N 1621
## 2 Y 98
## Popular.fctr.predict.Conditional.X.no.rnorm.rf.Y
## 1 92
## 2 246
## Reference
## Prediction N Y
## N 1621 98
## Y 92 246
## Popular.fctr Popular.fctr.predict.Conditional.X.no.rnorm.rf.N
## 1 N 1621
## 2 Y 98
## Popular.fctr.predict.Conditional.X.no.rnorm.rf.Y
## 1 92
## 2 246
## [1] "in mypredict_mdl: calling confusionMatrix..."
## [1] " unique(df[, rsp_var_out]: "
## [1] N Y
## Levels: N Y
## [1] " unique(df[, rsp_var ]: "
## [1] N Y
## Levels: N Y
## Prediction
## Reference N Y
## N 1621 92
## Y 98 246
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 9.076325e-01 6.660515e-01 8.942890e-01 9.198004e-01 8.327662e-01
## AccuracyPValue McnemarPValue
## 9.289451e-23 7.168005e-01
## model_id model_method
## 1 Conditional.X.no.rnorm.rf rf
## feats
## 1 WordCount, PubDate.hour, PubDate.month, PubDate.date, PubDate.second, PubDate.minute, PubDate.wkday, S.year, A.year, S.compani, A.compani, S.time, A.time, S.will, A.will, S.york, A.york, S.new, A.new, H.new, S.week, A.week, SectionName.fctr, NewsDesk.fctr, SubsectionName.fctr, H.num.chars, H.num.chars.log, A.num.chars, S.num.chars, H.num.words, H.num.words.unq, H.num.words.log, S.num.chars.log, A.num.chars.log, A.num.words, S.num.words, H.num.words.unq.log, A.num.words.unq, S.num.words.unq, A.num.words.log, S.num.words.log, A.num.words.unq.log, S.num.words.unq.log
## max.nTuningRuns min.elapsedtime.everything min.elapsedtime.final
## 1 3 100.692 25.241
## max.auc.fit opt.prob.threshold.fit max.f.score.fit max.Accuracy.fit
## 1 1 0.6 1 0.907486
## max.AccuracyLower.fit max.AccuracyUpper.fit max.Kappa.fit max.auc.OOB
## 1 0.999176 1 0.6429373 0.9283871
## opt.prob.threshold.OOB max.f.score.OOB max.Accuracy.OOB
## 1 0.4 0.7214076 0.9076325
## max.AccuracyLower.OOB max.AccuracyUpper.OOB max.Kappa.OOB
## 1 0.894289 0.9198004 0.6660515# User specified
# easier to exclude features
#model_id_pfx <- "";
# indep_vars_vctr <- setdiff(names(glb_fitent_df),
# union(union(glb_rsp_var, glb_exclude_vars_as_features),
# c("<feat1_name>", "<feat2_name>")))
# method <- ""
# easier to include features
#model_id_pfx <- ""; indep_vars_vctr <- c("<feat1_name>", "<feat1_name>"); method <- ""
# User specified bivariate models
# indep_vars_vctr_lst <- list()
# for (feat in setdiff(names(glb_fitent_df),
# union(glb_rsp_var, glb_exclude_vars_as_features)))
# indep_vars_vctr_lst[["feat"]] <- feat
# User specified combinatorial models
# indep_vars_vctr_lst <- list()
# combn_mtrx <- combn(c("<feat1_name>", "<feat2_name>", "<featn_name>"),
# <num_feats_to_choose>)
# for (combn_ix in 1:ncol(combn_mtrx))
# #print(combn_mtrx[, combn_ix])
# indep_vars_vctr_lst[[combn_ix]] <- combn_mtrx[, combn_ix]
# template for myfit_mdl
# rf is hard-coded in caret to recognize only Accuracy / Kappa evaluation metrics
# only for OOB in trainControl ?
# ret_lst <- myfit_mdl_fn(model_id=paste0(model_id_pfx, ""), model_method=method,
# indep_vars_vctr=indep_vars_vctr,
# rsp_var=glb_rsp_var, rsp_var_out=glb_rsp_var_out,
# fit_df=glb_fitent_df, OOB_df=glb_OOBent_df,
# n_cv_folds=glb_n_cv_folds, tune_models_df=glb_tune_models_df,
# model_loss_mtrx=glb_model_metric_terms,
# model_summaryFunction=glb_model_metric_smmry,
# model_metric=glb_model_metric,
# model_metric_maximize=glb_model_metric_maximize)
# Simplify a model
# fit_df <- glb_fitent_df; glb_mdl <- step(<complex>_mdl)
# Non-caret models
# rpart_area_mdl <- rpart(reformulate("Area", response=glb_rsp_var),
# data=glb_fitent_df, #method="class",
# control=rpart.control(cp=0.12),
# parms=list(loss=glb_model_metric_terms))
# print("rpart_sel_wlm_mdl"); prp(rpart_sel_wlm_mdl)
#
print(glb_models_df)## model_id model_method
## 1 MFO.myMFO_classfr myMFO_classfr
## 2 Random.myrandom_classfr myrandom_classfr
## 3 Max.cor.Y.cv.0.rpart rpart
## 4 Max.cor.Y.cv.0.cp.0.rpart rpart
## 5 Max.cor.Y.rpart rpart
## 6 Max.cor.Y.glm glm
## 7 Interact.High.cor.Y.glm glm
## 8 Low.cor.X.glm glm
## 9 Conditional.X.glm glm
## 10 Conditional.X.rpart rpart
## 11 Conditional.X.cp.0.rpart rpart
## 12 Conditional.X.rf rf
## 13 Conditional.X.no.rnorm.rf rf
## feats
## 1 .rnorm
## 2 .rnorm
## 3 WordCount
## 4 WordCount
## 5 WordCount
## 6 WordCount
## 7 WordCount, WordCount:S.year, WordCount:S.compani, WordCount:S.time, WordCount:S.will, WordCount:S.york, WordCount:S.new, WordCount:S.week, WordCount:H.num.chars, WordCount:A.num.chars, WordCount:H.num.words, WordCount:S.num.chars.log, WordCount:A.num.words, WordCount:H.num.chars.log, WordCount:S.num.words, WordCount:A.num.words.log, WordCount:S.num.chars
## 8 WordCount, PubDate.hour, PubDate.month, .rnorm, PubDate.date, PubDate.second, PubDate.minute, PubDate.wkday, S.year, S.compani, S.time, S.will, S.york, S.new, H.new, S.week, SectionName.fctr, NewsDesk.fctr, SubsectionName.fctr, H.num.chars, A.num.chars, H.num.words, H.num.words.log, S.num.chars.log, A.num.words, A.num.words.unq, A.num.words.log, A.num.words.unq.log
## 9 WordCount, PubDate.hour, PubDate.month, .rnorm, PubDate.date, PubDate.second, PubDate.minute, PubDate.wkday, S.year, A.year, S.compani, A.compani, S.time, A.time, S.will, A.will, S.york, A.york, S.new, A.new, H.new, S.week, A.week, SectionName.fctr, NewsDesk.fctr, SubsectionName.fctr, H.num.chars, H.num.chars.log, A.num.chars, S.num.chars, H.num.words, H.num.words.unq, H.num.words.log, S.num.chars.log, A.num.chars.log, A.num.words, S.num.words, H.num.words.unq.log, A.num.words.unq, S.num.words.unq, A.num.words.log, S.num.words.log, A.num.words.unq.log, S.num.words.unq.log
## 10 WordCount, PubDate.hour, PubDate.month, PubDate.date, PubDate.second, PubDate.minute, PubDate.wkday, S.year, A.year, S.compani, A.compani, S.time, A.time, S.will, A.will, S.york, A.york, S.new, A.new, H.new, S.week, A.week, SectionName.fctr, NewsDesk.fctr, SubsectionName.fctr, H.num.chars, H.num.chars.log, A.num.chars, S.num.chars, H.num.words, H.num.words.unq, H.num.words.log, S.num.chars.log, A.num.chars.log, A.num.words, S.num.words, H.num.words.unq.log, A.num.words.unq, S.num.words.unq, A.num.words.log, S.num.words.log, A.num.words.unq.log, S.num.words.unq.log
## 11 WordCount, PubDate.hour, PubDate.month, PubDate.date, PubDate.second, PubDate.minute, PubDate.wkday, S.year, A.year, S.compani, A.compani, S.time, A.time, S.will, A.will, S.york, A.york, S.new, A.new, H.new, S.week, A.week, SectionName.fctr, NewsDesk.fctr, SubsectionName.fctr, H.num.chars, H.num.chars.log, A.num.chars, S.num.chars, H.num.words, H.num.words.unq, H.num.words.log, S.num.chars.log, A.num.chars.log, A.num.words, S.num.words, H.num.words.unq.log, A.num.words.unq, S.num.words.unq, A.num.words.log, S.num.words.log, A.num.words.unq.log, S.num.words.unq.log
## 12 WordCount, PubDate.hour, PubDate.month, .rnorm, PubDate.date, PubDate.second, PubDate.minute, PubDate.wkday, S.year, A.year, S.compani, A.compani, S.time, A.time, S.will, A.will, S.york, A.york, S.new, A.new, H.new, S.week, A.week, SectionName.fctr, NewsDesk.fctr, SubsectionName.fctr, H.num.chars, H.num.chars.log, A.num.chars, S.num.chars, H.num.words, H.num.words.unq, H.num.words.log, S.num.chars.log, A.num.chars.log, A.num.words, S.num.words, H.num.words.unq.log, A.num.words.unq, S.num.words.unq, A.num.words.log, S.num.words.log, A.num.words.unq.log, S.num.words.unq.log
## 13 WordCount, PubDate.hour, PubDate.month, PubDate.date, PubDate.second, PubDate.minute, PubDate.wkday, S.year, A.year, S.compani, A.compani, S.time, A.time, S.will, A.will, S.york, A.york, S.new, A.new, H.new, S.week, A.week, SectionName.fctr, NewsDesk.fctr, SubsectionName.fctr, H.num.chars, H.num.chars.log, A.num.chars, S.num.chars, H.num.words, H.num.words.unq, H.num.words.log, S.num.chars.log, A.num.chars.log, A.num.words, S.num.words, H.num.words.unq.log, A.num.words.unq, S.num.words.unq, A.num.words.log, S.num.words.log, A.num.words.unq.log, S.num.words.unq.log
## max.nTuningRuns min.elapsedtime.everything min.elapsedtime.final
## 1 0 0.460 0.003
## 2 0 0.336 0.002
## 3 0 0.753 0.063
## 4 0 0.599 0.054
## 5 3 1.265 0.063
## 6 1 1.126 0.062
## 7 1 1.834 0.278
## 8 1 4.791 1.252
## 9 1 6.841 1.746
## 10 3 5.220 0.912
## 11 0 1.525 0.897
## 12 3 105.229 25.835
## 13 3 100.692 25.241
## max.auc.fit opt.prob.threshold.fit max.f.score.fit max.Accuracy.fit
## 1 0.5000000 0.5 0.0000000 0.8326257
## 2 0.4975446 0.1 0.2867534 0.1673743
## 3 0.5000000 0.5 0.0000000 0.8326257
## 4 0.7072558 0.2 0.4562462 0.7986592
## 5 0.5000000 0.5 0.0000000 0.8192183
## 6 0.7350785 0.2 0.4220705 0.8225702
## 7 0.8062116 0.2 0.5124688 0.8359773
## 8 0.9372948 0.3 0.7191737 0.9003357
## 9 0.9381279 0.3 0.7198970 0.8992181
## 10 0.8300780 0.7 0.5738832 0.8893849
## 11 0.9468583 0.4 0.7953930 0.9325140
## 12 1.0000000 0.5 1.0000000 0.9086034
## 13 1.0000000 0.6 1.0000000 0.9074860
## max.AccuracyLower.fit max.AccuracyUpper.fit max.Kappa.fit max.auc.OOB
## 1 0.8213602 0.8434553 0.00000000 0.5000000
## 2 0.1565447 0.1786398 0.00000000 0.4821958
## 3 0.8213602 0.8434553 0.00000000 0.5000000
## 4 0.7866053 0.8103220 0.33409619 0.6552144
## 5 0.8213602 0.8434553 0.07367080 0.5000000
## 6 0.7603993 0.7851652 0.01764536 0.7378291
## 7 0.8138022 0.8362745 0.16186063 0.8156980
## 8 0.8937373 0.9113196 0.61968400 0.9326067
## 9 0.8937373 0.9113196 0.61736565 0.9327051
## 10 0.8795973 0.8982154 0.52225926 0.8225030
## 11 0.9247659 0.9396915 0.75499696 0.8910062
## 12 0.9991760 1.0000000 0.64785575 0.9287358
## 13 0.9991760 1.0000000 0.64293726 0.9283871
## opt.prob.threshold.OOB max.f.score.OOB max.Accuracy.OOB
## 1 0.5 0.0000000 0.8327662
## 2 0.1 0.2865473 0.1672338
## 3 0.5 0.0000000 0.8327662
## 4 0.1 0.4058625 0.7438017
## 5 0.5 0.0000000 0.8327662
## 6 0.2 0.4044944 0.7681089
## 7 0.2 0.5000000 0.8142927
## 8 0.3 0.7136929 0.8993680
## 9 0.3 0.7081604 0.8974234
## 10 0.7 0.6014493 0.8930481
## 11 0.8 0.6563574 0.9027710
## 12 0.4 0.7195301 0.9071463
## 13 0.4 0.7214076 0.9076325
## max.AccuracyLower.OOB max.AccuracyUpper.OOB max.Kappa.OOB
## 1 0.8159247 0.8486533 0.0000000
## 2 0.1513467 0.1840753 0.0000000
## 3 0.8159247 0.8486533 0.0000000
## 4 0.7243549 0.7625511 0.2528893
## 5 0.8159247 0.8486533 0.0000000
## 6 0.7492527 0.7861979 0.2640573
## 7 0.7968011 0.8308835 0.3873528
## 8 0.8855497 0.9120357 0.6528225
## 9 0.8834974 0.9102047 0.6461137
## 10 0.8788850 0.9060795 0.5439760
## 11 0.8891448 0.9152364 0.6019082
## 12 0.8937741 0.9193445 0.6639007
## 13 0.8942890 0.9198004 0.6660515
## max.AccuracySD.fit max.KappaSD.fit min.aic.fit
## 1 NA NA NA
## 2 NA NA NA
## 3 NA NA NA
## 4 NA NA NA
## 5 0.003061652 0.025245020 NA
## 6 0.003663409 0.012343629 3778.810
## 7 0.003949821 0.031367400 3582.118
## 8 0.002679281 0.006873523 2206.319
## 9 0.002140938 0.001775530 2212.301
## 10 0.002458760 0.005834195 NA
## 11 NA NA NA
## 12 NA NA NA
## 13 NA NA NAglb_script_df <- rbind(glb_script_df,
data.frame(chunk_label="fit.models",
chunk_step_major=glb_script_df[nrow(glb_script_df), "chunk_step_major"],
chunk_step_minor=glb_script_df[nrow(glb_script_df), "chunk_step_minor"]+1,
elapsed=(proc.time() - glb_script_tm)["elapsed"]))
print(tail(glb_script_df, 2))## chunk_label chunk_step_major chunk_step_minor elapsed
## elapsed10 fit.models 6 1 180.884
## elapsed11 fit.models 6 2 433.449if (!is.null(glb_model_metric_smmry)) {
stats_df <- glb_models_df[, "model_id", FALSE]
stats_mdl_df <- data.frame()
for (model_id in stats_df$model_id) {
stats_mdl_df <- rbind(stats_mdl_df,
mypredict_mdl(glb_models_lst[[model_id]], glb_fitent_df, glb_rsp_var,
glb_rsp_var_out, model_id, "fit",
glb_model_metric_smmry, glb_model_metric,
glb_model_metric_maximize, ret_type="stats"))
}
stats_df <- merge(stats_df, stats_mdl_df, all.x=TRUE)
stats_mdl_df <- data.frame()
for (model_id in stats_df$model_id) {
stats_mdl_df <- rbind(stats_mdl_df,
mypredict_mdl(glb_models_lst[[model_id]], glb_OOBent_df, glb_rsp_var,
glb_rsp_var_out, model_id, "OOB",
glb_model_metric_smmry, glb_model_metric,
glb_model_metric_maximize, ret_type="stats"))
}
stats_df <- merge(stats_df, stats_mdl_df, all.x=TRUE)
# tmp_models_df <- orderBy(~model_id, glb_models_df)
# rownames(tmp_models_df) <- seq(1, nrow(tmp_models_df))
# all.equal(subset(tmp_models_df[, names(stats_df)], model_id != "Random.myrandom_classfr"),
# subset(stats_df, model_id != "Random.myrandom_classfr"))
# print(subset(tmp_models_df[, names(stats_df)], model_id != "Random.myrandom_classfr")[, c("model_id", "max.Accuracy.fit")])
# print(subset(stats_df, model_id != "Random.myrandom_classfr")[, c("model_id", "max.Accuracy.fit")])
print("Merging following data into glb_models_df:")
print(stats_mrg_df <- stats_df[, c(1, grep(glb_model_metric, names(stats_df)))])
print(tmp_models_df <- orderBy(~model_id, glb_models_df[, c("model_id", grep(glb_model_metric, names(stats_df), value=TRUE))]))
tmp2_models_df <- glb_models_df[, c("model_id", setdiff(names(glb_models_df), grep(glb_model_metric, names(stats_df), value=TRUE)))]
tmp3_models_df <- merge(tmp2_models_df, stats_mrg_df, all.x=TRUE, sort=FALSE)
print(tmp3_models_df)
print(names(tmp3_models_df))
print(glb_models_df <- subset(tmp3_models_df, select=-model_id.1))
}
plt_models_df <- glb_models_df[, -grep("SD|Upper|Lower", names(glb_models_df))]
for (var in grep("^min.", names(plt_models_df), value=TRUE)) {
plt_models_df[, sub("min.", "inv.", var)] <-
#ifelse(all(is.na(tmp <- plt_models_df[, var])), NA, 1.0 / tmp)
1.0 / plt_models_df[, var]
plt_models_df <- plt_models_df[ , -grep(var, names(plt_models_df))]
}
print(plt_models_df)## model_id model_method
## 1 MFO.myMFO_classfr myMFO_classfr
## 2 Random.myrandom_classfr myrandom_classfr
## 3 Max.cor.Y.cv.0.rpart rpart
## 4 Max.cor.Y.cv.0.cp.0.rpart rpart
## 5 Max.cor.Y.rpart rpart
## 6 Max.cor.Y.glm glm
## 7 Interact.High.cor.Y.glm glm
## 8 Low.cor.X.glm glm
## 9 Conditional.X.glm glm
## 10 Conditional.X.rpart rpart
## 11 Conditional.X.cp.0.rpart rpart
## 12 Conditional.X.rf rf
## 13 Conditional.X.no.rnorm.rf rf
## feats
## 1 .rnorm
## 2 .rnorm
## 3 WordCount
## 4 WordCount
## 5 WordCount
## 6 WordCount
## 7 WordCount, WordCount:S.year, WordCount:S.compani, WordCount:S.time, WordCount:S.will, WordCount:S.york, WordCount:S.new, WordCount:S.week, WordCount:H.num.chars, WordCount:A.num.chars, WordCount:H.num.words, WordCount:S.num.chars.log, WordCount:A.num.words, WordCount:H.num.chars.log, WordCount:S.num.words, WordCount:A.num.words.log, WordCount:S.num.chars
## 8 WordCount, PubDate.hour, PubDate.month, .rnorm, PubDate.date, PubDate.second, PubDate.minute, PubDate.wkday, S.year, S.compani, S.time, S.will, S.york, S.new, H.new, S.week, SectionName.fctr, NewsDesk.fctr, SubsectionName.fctr, H.num.chars, A.num.chars, H.num.words, H.num.words.log, S.num.chars.log, A.num.words, A.num.words.unq, A.num.words.log, A.num.words.unq.log
## 9 WordCount, PubDate.hour, PubDate.month, .rnorm, PubDate.date, PubDate.second, PubDate.minute, PubDate.wkday, S.year, A.year, S.compani, A.compani, S.time, A.time, S.will, A.will, S.york, A.york, S.new, A.new, H.new, S.week, A.week, SectionName.fctr, NewsDesk.fctr, SubsectionName.fctr, H.num.chars, H.num.chars.log, A.num.chars, S.num.chars, H.num.words, H.num.words.unq, H.num.words.log, S.num.chars.log, A.num.chars.log, A.num.words, S.num.words, H.num.words.unq.log, A.num.words.unq, S.num.words.unq, A.num.words.log, S.num.words.log, A.num.words.unq.log, S.num.words.unq.log
## 10 WordCount, PubDate.hour, PubDate.month, PubDate.date, PubDate.second, PubDate.minute, PubDate.wkday, S.year, A.year, S.compani, A.compani, S.time, A.time, S.will, A.will, S.york, A.york, S.new, A.new, H.new, S.week, A.week, SectionName.fctr, NewsDesk.fctr, SubsectionName.fctr, H.num.chars, H.num.chars.log, A.num.chars, S.num.chars, H.num.words, H.num.words.unq, H.num.words.log, S.num.chars.log, A.num.chars.log, A.num.words, S.num.words, H.num.words.unq.log, A.num.words.unq, S.num.words.unq, A.num.words.log, S.num.words.log, A.num.words.unq.log, S.num.words.unq.log
## 11 WordCount, PubDate.hour, PubDate.month, PubDate.date, PubDate.second, PubDate.minute, PubDate.wkday, S.year, A.year, S.compani, A.compani, S.time, A.time, S.will, A.will, S.york, A.york, S.new, A.new, H.new, S.week, A.week, SectionName.fctr, NewsDesk.fctr, SubsectionName.fctr, H.num.chars, H.num.chars.log, A.num.chars, S.num.chars, H.num.words, H.num.words.unq, H.num.words.log, S.num.chars.log, A.num.chars.log, A.num.words, S.num.words, H.num.words.unq.log, A.num.words.unq, S.num.words.unq, A.num.words.log, S.num.words.log, A.num.words.unq.log, S.num.words.unq.log
## 12 WordCount, PubDate.hour, PubDate.month, .rnorm, PubDate.date, PubDate.second, PubDate.minute, PubDate.wkday, S.year, A.year, S.compani, A.compani, S.time, A.time, S.will, A.will, S.york, A.york, S.new, A.new, H.new, S.week, A.week, SectionName.fctr, NewsDesk.fctr, SubsectionName.fctr, H.num.chars, H.num.chars.log, A.num.chars, S.num.chars, H.num.words, H.num.words.unq, H.num.words.log, S.num.chars.log, A.num.chars.log, A.num.words, S.num.words, H.num.words.unq.log, A.num.words.unq, S.num.words.unq, A.num.words.log, S.num.words.log, A.num.words.unq.log, S.num.words.unq.log
## 13 WordCount, PubDate.hour, PubDate.month, PubDate.date, PubDate.second, PubDate.minute, PubDate.wkday, S.year, A.year, S.compani, A.compani, S.time, A.time, S.will, A.will, S.york, A.york, S.new, A.new, H.new, S.week, A.week, SectionName.fctr, NewsDesk.fctr, SubsectionName.fctr, H.num.chars, H.num.chars.log, A.num.chars, S.num.chars, H.num.words, H.num.words.unq, H.num.words.log, S.num.chars.log, A.num.chars.log, A.num.words, S.num.words, H.num.words.unq.log, A.num.words.unq, S.num.words.unq, A.num.words.log, S.num.words.log, A.num.words.unq.log, S.num.words.unq.log
## max.nTuningRuns max.auc.fit opt.prob.threshold.fit max.f.score.fit
## 1 0 0.5000000 0.5 0.0000000
## 2 0 0.4975446 0.1 0.2867534
## 3 0 0.5000000 0.5 0.0000000
## 4 0 0.7072558 0.2 0.4562462
## 5 3 0.5000000 0.5 0.0000000
## 6 1 0.7350785 0.2 0.4220705
## 7 1 0.8062116 0.2 0.5124688
## 8 1 0.9372948 0.3 0.7191737
## 9 1 0.9381279 0.3 0.7198970
## 10 3 0.8300780 0.7 0.5738832
## 11 0 0.9468583 0.4 0.7953930
## 12 3 1.0000000 0.5 1.0000000
## 13 3 1.0000000 0.6 1.0000000
## max.Accuracy.fit max.Kappa.fit max.auc.OOB opt.prob.threshold.OOB
## 1 0.8326257 0.00000000 0.5000000 0.5
## 2 0.1673743 0.00000000 0.4821958 0.1
## 3 0.8326257 0.00000000 0.5000000 0.5
## 4 0.7986592 0.33409619 0.6552144 0.1
## 5 0.8192183 0.07367080 0.5000000 0.5
## 6 0.8225702 0.01764536 0.7378291 0.2
## 7 0.8359773 0.16186063 0.8156980 0.2
## 8 0.9003357 0.61968400 0.9326067 0.3
## 9 0.8992181 0.61736565 0.9327051 0.3
## 10 0.8893849 0.52225926 0.8225030 0.7
## 11 0.9325140 0.75499696 0.8910062 0.8
## 12 0.9086034 0.64785575 0.9287358 0.4
## 13 0.9074860 0.64293726 0.9283871 0.4
## max.f.score.OOB max.Accuracy.OOB max.Kappa.OOB
## 1 0.0000000 0.8327662 0.0000000
## 2 0.2865473 0.1672338 0.0000000
## 3 0.0000000 0.8327662 0.0000000
## 4 0.4058625 0.7438017 0.2528893
## 5 0.0000000 0.8327662 0.0000000
## 6 0.4044944 0.7681089 0.2640573
## 7 0.5000000 0.8142927 0.3873528
## 8 0.7136929 0.8993680 0.6528225
## 9 0.7081604 0.8974234 0.6461137
## 10 0.6014493 0.8930481 0.5439760
## 11 0.6563574 0.9027710 0.6019082
## 12 0.7195301 0.9071463 0.6639007
## 13 0.7214076 0.9076325 0.6660515
## inv.elapsedtime.everything inv.elapsedtime.final inv.aic.fit
## 1 2.173913043 333.33333333 NA
## 2 2.976190476 500.00000000 NA
## 3 1.328021248 15.87301587 NA
## 4 1.669449082 18.51851852 NA
## 5 0.790513834 15.87301587 NA
## 6 0.888099467 16.12903226 0.0002646336
## 7 0.545256270 3.59712230 0.0002791645
## 8 0.208724692 0.79872204 0.0004532437
## 9 0.146177459 0.57273769 0.0004520180
## 10 0.191570881 1.09649123 NA
## 11 0.655737705 1.11482720 NA
## 12 0.009503084 0.03870718 NA
## 13 0.009931276 0.03961808 NAprint(myplot_radar(radar_inp_df=plt_models_df))## Warning in RColorBrewer::brewer.pal(n, pal): n too large, allowed maximum for palette Set1 is 9
## Returning the palette you asked for with that many colors## Warning: The shape palette can deal with a maximum of 6 discrete values
## because more than 6 becomes difficult to discriminate; you have
## 13. Consider specifying shapes manually. if you must have them.## Warning: Removed 6 rows containing missing values (geom_path).## Warning: Removed 104 rows containing missing values (geom_point).## Warning: Removed 9 rows containing missing values (geom_text).## Warning in RColorBrewer::brewer.pal(n, pal): n too large, allowed maximum for palette Set1 is 9
## Returning the palette you asked for with that many colors## Warning: The shape palette can deal with a maximum of 6 discrete values
## because more than 6 becomes difficult to discriminate; you have
## 13. Consider specifying shapes manually. if you must have them.
# print(myplot_radar(radar_inp_df=subset(plt_models_df,
# !(model_id %in% grep("random|MFO", plt_models_df$model_id, value=TRUE)))))
# Compute CI for <metric>SD
glb_models_df <- mutate(glb_models_df,
max.df = ifelse(max.nTuningRuns > 1, max.nTuningRuns - 1, NA),
min.sd2ci.scaler = ifelse(is.na(max.df), NA, qt(0.975, max.df)))
for (var in grep("SD", names(glb_models_df), value=TRUE)) {
# Does CI alredy exist ?
var_components <- unlist(strsplit(var, "SD"))
varActul <- paste0(var_components[1], var_components[2])
varUpper <- paste0(var_components[1], "Upper", var_components[2])
varLower <- paste0(var_components[1], "Lower", var_components[2])
if (varUpper %in% names(glb_models_df)) {
warning(varUpper, " already exists in glb_models_df")
# Assuming Lower also exists
next
}
print(sprintf("var:%s", var))
# CI is dependent on sample size in t distribution; df=n-1
glb_models_df[, varUpper] <- glb_models_df[, varActul] +
glb_models_df[, "min.sd2ci.scaler"] * glb_models_df[, var]
glb_models_df[, varLower] <- glb_models_df[, varActul] -
glb_models_df[, "min.sd2ci.scaler"] * glb_models_df[, var]
}## Warning: max.AccuracyUpper.fit already exists in glb_models_df## [1] "var:max.KappaSD.fit"# Plot metrics with CI
plt_models_df <- glb_models_df[, "model_id", FALSE]
pltCI_models_df <- glb_models_df[, "model_id", FALSE]
for (var in grep("Upper", names(glb_models_df), value=TRUE)) {
var_components <- unlist(strsplit(var, "Upper"))
col_name <- unlist(paste(var_components, collapse=""))
plt_models_df[, col_name] <- glb_models_df[, col_name]
for (name in paste0(var_components[1], c("Upper", "Lower"), var_components[2]))
pltCI_models_df[, name] <- glb_models_df[, name]
}
build_statsCI_data <- function(plt_models_df) {
mltd_models_df <- melt(plt_models_df, id.vars="model_id")
mltd_models_df$data <- sapply(1:nrow(mltd_models_df),
function(row_ix) tail(unlist(strsplit(as.character(
mltd_models_df[row_ix, "variable"]), "[.]")), 1))
mltd_models_df$label <- sapply(1:nrow(mltd_models_df),
function(row_ix) head(unlist(strsplit(as.character(
mltd_models_df[row_ix, "variable"]), paste0(".", mltd_models_df[row_ix, "data"]))), 1))
#print(mltd_models_df)
return(mltd_models_df)
}
mltd_models_df <- build_statsCI_data(plt_models_df)
mltdCI_models_df <- melt(pltCI_models_df, id.vars="model_id")
for (row_ix in 1:nrow(mltdCI_models_df)) {
for (type in c("Upper", "Lower")) {
if (length(var_components <- unlist(strsplit(
as.character(mltdCI_models_df[row_ix, "variable"]), type))) > 1) {
#print(sprintf("row_ix:%d; type:%s; ", row_ix, type))
mltdCI_models_df[row_ix, "label"] <- var_components[1]
mltdCI_models_df[row_ix, "data"] <- unlist(strsplit(var_components[2], "[.]"))[2]
mltdCI_models_df[row_ix, "type"] <- type
break
}
}
}
#print(mltdCI_models_df)
# castCI_models_df <- dcast(mltdCI_models_df, value ~ type, fun.aggregate=sum)
# print(castCI_models_df)
wideCI_models_df <- reshape(subset(mltdCI_models_df, select=-variable),
timevar="type",
idvar=setdiff(names(mltdCI_models_df), c("type", "value", "variable")),
direction="wide")
#print(wideCI_models_df)
mrgdCI_models_df <- merge(wideCI_models_df, mltd_models_df, all.x=TRUE)
#print(mrgdCI_models_df)
# Merge stats back in if CIs don't exist
goback_vars <- c()
for (var in unique(mltd_models_df$label)) {
for (type in unique(mltd_models_df$data)) {
var_type <- paste0(var, ".", type)
# if this data is already present, next
if (var_type %in% unique(paste(mltd_models_df$label, mltd_models_df$data, sep=".")))
next
#print(sprintf("var_type:%s", var_type))
goback_vars <- c(goback_vars, var_type)
}
}
if (length(goback_vars) > 0) {
mltd_goback_df <- build_statsCI_data(glb_models_df[, c("model_id", goback_vars)])
mltd_models_df <- rbind(mltd_models_df, mltd_goback_df)
}
mltd_models_df <- merge(mltd_models_df, glb_models_df[, c("model_id", "model_method")], all.x=TRUE)
png(paste0(glb_out_pfx, "models_bar.png"), width=480*3, height=480*2)
print(gp <- myplot_bar(mltd_models_df, "model_id", "value", colorcol_name="model_method") +
geom_errorbar(data=mrgdCI_models_df,
mapping=aes(x=model_id, ymax=value.Upper, ymin=value.Lower), width=0.5) +
facet_grid(label ~ data, scales="free") +
theme(axis.text.x = element_text(angle = 90,vjust = 0.5)))
dev.off()## quartz_off_screen
## 2print(gp)
# used for console inspection
model_evl_terms <- c(NULL)
for (metric in glb_model_evl_criteria)
model_evl_terms <- c(model_evl_terms,
ifelse(length(grep("max", metric)) > 0, "-", "+"), metric)
model_sel_frmla <- as.formula(paste(c("~ ", model_evl_terms), collapse=" "))
print(tmp_models_df <- orderBy(model_sel_frmla, glb_models_df)[, c("model_id", glb_model_evl_criteria)])## model_id max.Accuracy.OOB max.Kappa.OOB min.aic.fit
## 13 Conditional.X.no.rnorm.rf 0.9076325 0.6660515 NA
## 12 Conditional.X.rf 0.9071463 0.6639007 NA
## 11 Conditional.X.cp.0.rpart 0.9027710 0.6019082 NA
## 8 Low.cor.X.glm 0.8993680 0.6528225 2206.319
## 9 Conditional.X.glm 0.8974234 0.6461137 2212.301
## 10 Conditional.X.rpart 0.8930481 0.5439760 NA
## 1 MFO.myMFO_classfr 0.8327662 0.0000000 NA
## 3 Max.cor.Y.cv.0.rpart 0.8327662 0.0000000 NA
## 5 Max.cor.Y.rpart 0.8327662 0.0000000 NA
## 7 Interact.High.cor.Y.glm 0.8142927 0.3873528 3582.118
## 6 Max.cor.Y.glm 0.7681089 0.2640573 3778.810
## 4 Max.cor.Y.cv.0.cp.0.rpart 0.7438017 0.2528893 NA
## 2 Random.myrandom_classfr 0.1672338 0.0000000 NAprint(myplot_radar(radar_inp_df=tmp_models_df))## Warning in RColorBrewer::brewer.pal(n, pal): n too large, allowed maximum for palette Set1 is 9
## Returning the palette you asked for with that many colors## Warning: The shape palette can deal with a maximum of 6 discrete values
## because more than 6 becomes difficult to discriminate; you have
## 13. Consider specifying shapes manually. if you must have them.## Warning: Removed 6 rows containing missing values (geom_path).## Warning: Removed 27 rows containing missing values (geom_point).## Warning: Removed 9 rows containing missing values (geom_text).## Warning in RColorBrewer::brewer.pal(n, pal): n too large, allowed maximum for palette Set1 is 9
## Returning the palette you asked for with that many colors## Warning: The shape palette can deal with a maximum of 6 discrete values
## because more than 6 becomes difficult to discriminate; you have
## 13. Consider specifying shapes manually. if you must have them.
print("Metrics used for model selection:"); print(model_sel_frmla)## [1] "Metrics used for model selection:"## ~-max.Accuracy.OOB - max.Kappa.OOB + min.aic.fitprint(sprintf("Best model id: %s", tmp_models_df[1, "model_id"]))## [1] "Best model id: Conditional.X.no.rnorm.rf"if (is.null(glb_sel_mdl_id))
{ glb_sel_mdl_id <- tmp_models_df[1, "model_id"] } else
print(sprintf("User specified selection: %s", glb_sel_mdl_id))
myprint_mdl(glb_sel_mdl <- glb_models_lst[[glb_sel_mdl_id]])
## Length Class Mode
## call 4 -none- call
## type 1 -none- character
## predicted 4475 factor numeric
## err.rate 1500 -none- numeric
## confusion 6 -none- numeric
## votes 8950 matrix numeric
## oob.times 4475 -none- numeric
## classes 2 -none- character
## importance 75 -none- numeric
## importanceSD 0 -none- NULL
## localImportance 0 -none- NULL
## proximity 0 -none- NULL
## ntree 1 -none- numeric
## mtry 1 -none- numeric
## forest 14 -none- list
## y 4475 factor numeric
## test 0 -none- NULL
## inbag 0 -none- NULL
## xNames 75 -none- character
## problemType 1 -none- character
## tuneValue 1 data.frame list
## obsLevels 2 -none- character## [1] TRUEreplay.petrisim(pn=glb_analytics_pn,
replay.trans=(glb_analytics_avl_objs <- c(glb_analytics_avl_objs,
"model.selected")), flip_coord=TRUE)## time trans "bgn " "fit.data.training.all " "predict.data.new " "end "
## 0.0000 multiple enabled transitions: data.training.all data.new model.selected firing: data.training.all
## 1.0000 1 2 1 0 0
## 1.0000 multiple enabled transitions: data.training.all data.new model.selected model.final data.training.all.prediction firing: data.new
## 2.0000 2 1 1 1 0
## 2.0000 multiple enabled transitions: data.training.all data.new model.selected model.final data.training.all.prediction data.new.prediction firing: model.selected
## 3.0000 3 0 2 1 0
glb_script_df <- rbind(glb_script_df,
data.frame(chunk_label="fit.data.training",
chunk_step_major=max(glb_script_df$chunk_step_major)+1,
chunk_step_minor=0,
elapsed=(proc.time() - glb_script_tm)["elapsed"]))
print(tail(glb_script_df, 2))## chunk_label chunk_step_major chunk_step_minor elapsed
## elapsed11 fit.models 6 2 433.449
## elapsed12 fit.data.training 7 0 447.799Step 7: fit.data.training
if (!is.null(glb_fin_mdl_id) && (glb_fin_mdl_id %in% names(glb_models_lst))) {
warning("Final model same as user selected model")
glb_fin_mdl <- glb_sel_mdl
} else {
print(mdl_feats_df <- myextract_mdl_feats(sel_mdl=glb_sel_mdl,
entity_df=glb_fitent_df))
if ((model_method <- glb_sel_mdl$method) == "custom")
# get actual method from the model_id
model_method <- tail(unlist(strsplit(glb_sel_mdl_id, "[.]")), 1)
tune_finmdl_df <- NULL
if (nrow(glb_sel_mdl$bestTune) > 0) {
for (param in names(glb_sel_mdl$bestTune)) {
#print(sprintf("param: %s", param))
tune_finmdl_df <- rbind(tune_finmdl_df,
data.frame(parameter=param,
min=glb_sel_mdl$bestTune[1, param],
max=glb_sel_mdl$bestTune[1, param],
by=1)) # by val does not matter
}
}
# Sync with parameters in mydsutils.R
ret_lst <- myfit_mdl(model_id="Final", model_method=model_method,
indep_vars_vctr=mdl_feats_df$id, model_type=glb_model_type,
rsp_var=glb_rsp_var, rsp_var_out=glb_rsp_var_out,
fit_df=glb_trnent_df, OOB_df=NULL,
n_cv_folds=glb_n_cv_folds, tune_models_df=tune_finmdl_df,
# Automate from here
# Issues if glb_sel_mdl$method == "rf" b/c trainControl is "oob"; not "cv"
model_loss_mtrx=glb_model_metric_terms,
model_summaryFunction=glb_sel_mdl$control$summaryFunction,
model_metric=glb_sel_mdl$metric,
model_metric_maximize=glb_sel_mdl$maximize)
glb_fin_mdl <- glb_models_lst[[length(glb_models_lst)]]
glb_fin_mdl_id <- glb_models_df[length(glb_models_lst), "model_id"]
}## id importance
## NewsDesk.fctr1 NewsDesk.fctr 100.0000000
## WordCount WordCount 92.2672600
## PubDate.hour PubDate.hour 45.0890753
## SectionName.fctr1 SectionName.fctr 38.4932377
## PubDate.minute PubDate.minute 28.0970381
## PubDate.second PubDate.second 23.6097109
## H.num.chars H.num.chars 20.4791833
## H.num.chars.log H.num.chars.log 20.3585154
## PubDate.date PubDate.date 20.0148418
## A.num.chars.log A.num.chars.log 17.7059570
## S.num.chars.log S.num.chars.log 17.1895057
## A.num.chars A.num.chars 17.0681326
## PubDate.wkday PubDate.wkday 17.0150952
## S.num.chars S.num.chars 16.9976533
## PubDate.month PubDate.month 7.4612116
## SectionName.fctr SectionName.fctr 7.3351215
## H.num.words H.num.words 6.5844434
## NewsDesk.fctr NewsDesk.fctr 6.5568902
## H.num.words.unq.log H.num.words.unq.log 6.4062094
## H.num.words.log H.num.words.log 6.3910724
## H.num.words.unq H.num.words.unq 6.3128616
## A.num.words.unq A.num.words.unq 5.3099741
## S.num.words.unq S.num.words.unq 5.2607201
## S.num.words.unq.log S.num.words.unq.log 5.1258698
## A.num.words.unq.log A.num.words.unq.log 5.1216312
## A.num.words A.num.words 4.9427359
## S.num.words S.num.words 4.7636974
## A.num.words.log A.num.words.log 4.7489550
## S.num.words.log S.num.words.log 4.6903604
## SubsectionName.fctr SubsectionName.fctr 4.2589466
## A.time A.time 1.7015104
## A.new A.new 1.6751538
## S.new S.new 1.6027706
## S.time S.time 1.4996134
## S.will S.will 1.4712946
## A.will A.will 1.4665449
## A.year A.year 1.3544952
## A.compani A.compani 1.3514317
## S.year S.year 1.3291077
## S.week S.week 1.2989863
## A.week A.week 1.2723817
## S.compani S.compani 1.2299062
## H.new H.new 0.9587475
## A.york A.york 0.4698017
## S.york S.york 0.4182270
## [1] "fitting model: Final.rf"
## [1] " indep_vars: NewsDesk.fctr, WordCount, PubDate.hour, SectionName.fctr, PubDate.minute, PubDate.second, H.num.chars, H.num.chars.log, PubDate.date, A.num.chars.log, S.num.chars.log, A.num.chars, PubDate.wkday, S.num.chars, PubDate.month, SectionName.fctr, H.num.words, NewsDesk.fctr, H.num.words.unq.log, H.num.words.log, H.num.words.unq, A.num.words.unq, S.num.words.unq, S.num.words.unq.log, A.num.words.unq.log, A.num.words, S.num.words, A.num.words.log, S.num.words.log, SubsectionName.fctr, A.time, A.new, S.new, S.time, S.will, A.will, A.year, A.compani, S.year, S.week, A.week, S.compani, H.new, A.york, S.york"
## + : mtry=38
## - : mtry=38
## Aggregating results
## Fitting final model on full training set
## Length Class Mode
## call 4 -none- call
## type 1 -none- character
## predicted 6532 factor numeric
## err.rate 1500 -none- numeric
## confusion 6 -none- numeric
## votes 13064 matrix numeric
## oob.times 6532 -none- numeric
## classes 2 -none- character
## importance 75 -none- numeric
## importanceSD 0 -none- NULL
## localImportance 0 -none- NULL
## proximity 0 -none- NULL
## ntree 1 -none- numeric
## mtry 1 -none- numeric
## forest 14 -none- list
## y 6532 factor numeric
## test 0 -none- NULL
## inbag 0 -none- NULL
## xNames 75 -none- character
## problemType 1 -none- character
## tuneValue 1 data.frame list
## obsLevels 2 -none- character
## [1] " calling mypredict_mdl for fit:"
## Reference
## Prediction N Y
## N 0 0
## Y 5439 1093
## Popular.fctr Popular.fctr.predict.Final.rf.N
## 1 N 0
## 2 Y 0
## Popular.fctr.predict.Final.rf.Y
## 1 5439
## 2 1093
## Reference
## Prediction N Y
## N 4882 0
## Y 557 1093
## Popular.fctr Popular.fctr.predict.Final.rf.N
## 1 N 4882
## 2 Y 0
## Popular.fctr.predict.Final.rf.Y
## 1 557
## 2 1093
## Reference
## Prediction N Y
## N 5278 0
## Y 161 1093
## Popular.fctr Popular.fctr.predict.Final.rf.N
## 1 N 5278
## 2 Y 0
## Popular.fctr.predict.Final.rf.Y
## 1 161
## 2 1093
## Reference
## Prediction N Y
## N 5392 0
## Y 47 1093
## Popular.fctr Popular.fctr.predict.Final.rf.N
## 1 N 5392
## 2 Y 0
## Popular.fctr.predict.Final.rf.Y
## 1 47
## 2 1093
## Reference
## Prediction N Y
## N 5439 0
## Y 0 1093
## Popular.fctr Popular.fctr.predict.Final.rf.N
## 1 N 5439
## 2 Y 0
## Popular.fctr.predict.Final.rf.Y
## 1 0
## 2 1093
## Reference
## Prediction N Y
## N 5439 0
## Y 0 1093
## Popular.fctr Popular.fctr.predict.Final.rf.N
## 1 N 5439
## 2 Y 0
## Popular.fctr.predict.Final.rf.Y
## 1 0
## 2 1093
## Reference
## Prediction N Y
## N 5439 3
## Y 0 1090
## Popular.fctr Popular.fctr.predict.Final.rf.N
## 1 N 5439
## 2 Y 3
## Popular.fctr.predict.Final.rf.Y
## 1 0
## 2 1090
## Reference
## Prediction N Y
## N 5439 151
## Y 0 942
## Popular.fctr Popular.fctr.predict.Final.rf.N
## 1 N 5439
## 2 Y 151
## Popular.fctr.predict.Final.rf.Y
## 1 0
## 2 942
## Reference
## Prediction N Y
## N 5439 350
## Y 0 743
## Popular.fctr Popular.fctr.predict.Final.rf.N
## 1 N 5439
## 2 Y 350
## Popular.fctr.predict.Final.rf.Y
## 1 0
## 2 743
## Reference
## Prediction N Y
## N 5439 602
## Y 0 491
## Popular.fctr Popular.fctr.predict.Final.rf.N
## 1 N 5439
## 2 Y 602
## Popular.fctr.predict.Final.rf.Y
## 1 0
## 2 491
## Reference
## Prediction N Y
## N 5439 1084
## Y 0 9
## Popular.fctr Popular.fctr.predict.Final.rf.N
## 1 N 5439
## 2 Y 1084
## Popular.fctr.predict.Final.rf.Y
## 1 0
## 2 9
## threshold f.score
## 1 0.0 0.28668852
## 2 0.1 0.79693766
## 3 0.2 0.93140179
## 4 0.3 0.97895208
## 5 0.4 1.00000000
## 6 0.5 1.00000000
## 7 0.6 0.99862574
## 8 0.7 0.92579853
## 9 0.8 0.80936819
## 10 0.9 0.61994949
## 11 1.0 0.01633394## [1] "Classifier Probability Threshold: 0.5000 to maximize f.score.fit"
## Popular.fctr Popular.fctr.predict.Final.rf.N
## 1 N 5439
## 2 Y NA
## Popular.fctr.predict.Final.rf.Y
## 1 NA
## 2 1093
## Reference
## Prediction N Y
## N 5439 0
## Y 0 1093
## Popular.fctr Popular.fctr.predict.Final.rf.N
## 1 N 5439
## 2 Y 0
## Popular.fctr.predict.Final.rf.Y
## 1 0
## 2 1093
## [1] "in mypredict_mdl: calling confusionMatrix..."
## [1] " unique(df[, rsp_var_out]: "
## [1] Y N
## Levels: N Y
## [1] " unique(df[, rsp_var ]: "
## [1] Y N
## Levels: N Y
## Prediction
## Reference N Y
## N 5439 0
## Y 0 1093
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 1.0000000 1.0000000 0.9994354 1.0000000 0.8326699
## AccuracyPValue McnemarPValue
## 0.0000000 NaN## Warning in mypredict_mdl(mdl, df = fit_df, rsp_var, rsp_var_out,
## model_id_method, : Expecting 1 metric: Accuracy; recd: Accuracy, Kappa;
## retaining Accuracy only
## model_id model_method
## 1 Final.rf rf
## feats
## 1 NewsDesk.fctr, WordCount, PubDate.hour, SectionName.fctr, PubDate.minute, PubDate.second, H.num.chars, H.num.chars.log, PubDate.date, A.num.chars.log, S.num.chars.log, A.num.chars, PubDate.wkday, S.num.chars, PubDate.month, SectionName.fctr, H.num.words, NewsDesk.fctr, H.num.words.unq.log, H.num.words.log, H.num.words.unq, A.num.words.unq, S.num.words.unq, S.num.words.unq.log, A.num.words.unq.log, A.num.words, S.num.words, A.num.words.log, S.num.words.log, SubsectionName.fctr, A.time, A.new, S.new, S.time, S.will, A.will, A.year, A.compani, S.year, S.week, A.week, S.compani, H.new, A.york, S.york
## max.nTuningRuns min.elapsedtime.everything min.elapsedtime.final
## 1 1 87.291 42.923
## max.auc.fit opt.prob.threshold.fit max.f.score.fit max.Accuracy.fit
## 1 1 0.5 1 0.9102878
## max.AccuracyLower.fit max.AccuracyUpper.fit max.Kappa.fit
## 1 0.9994354 1 0.6532154glb_script_df <- rbind(glb_script_df,
data.frame(chunk_label="fit.data.training",
chunk_step_major=glb_script_df[nrow(glb_script_df), "chunk_step_major"],
chunk_step_minor=glb_script_df[nrow(glb_script_df), "chunk_step_minor"]+1,
elapsed=(proc.time() - glb_script_tm)["elapsed"]))
print(tail(glb_script_df, 2))## chunk_label chunk_step_major chunk_step_minor elapsed
## elapsed12 fit.data.training 7 0 447.799
## elapsed13 fit.data.training 7 1 539.919glb_rsp_var_out <- paste0(glb_rsp_var_out, tail(names(glb_models_lst), 1))
# Used again in predict.data.new chunk
glb_get_predictions <- function(df) {
if (glb_is_regression) {
df[, glb_rsp_var_out] <- predict(glb_fin_mdl, newdata=df, type="raw")
print(myplot_scatter(df, glb_rsp_var, glb_rsp_var_out,
smooth=TRUE))
df[, paste0(glb_rsp_var_out, ".err")] <-
abs(df[, glb_rsp_var_out] - df[, glb_rsp_var])
print(head(orderBy(reformulate(c("-", paste0(glb_rsp_var_out, ".err"))),
df)))
}
if (glb_is_classification && glb_is_binomial) {
# incorporate glb_clf_proba_threshold
# shd it only be for glb_fin_mdl or for earlier models ?
# for glb_trnent_df it shd opt; vs. assume for glb_newent_df
finmdl_prob_fit <- glb_models_df[glb_models_df$model_id == glb_fin_mdl_id,
"opt.prob.threshold.fit"]
selmdl_prob_fit <- glb_models_df[glb_models_df$model_id == glb_sel_mdl_id,
"opt.prob.threshold.fit"]
if (finmdl_prob_fit != selmdl_prob_fit)
warning("opt.prob.threshold.fit differs for fin_mdl: ", finmdl_prob_fit,
" vs. sel_mdl: ", selmdl_prob_fit)
prob_threshold <- glb_models_df[glb_models_df$model_id == glb_sel_mdl_id,
"opt.prob.threshold.OOB"]
df[, paste0(glb_rsp_var_out, ".prob")] <-
predict(glb_fin_mdl, newdata=df, type="prob")[, 2]
df[, glb_rsp_var_out] <-
factor(levels(df[, glb_rsp_var])[
(df[, paste0(glb_rsp_var_out, ".prob")] >=
prob_threshold) * 1 + 1], levels(df[, glb_rsp_var]))
# prediction stats already reported by myfit_mdl ???
}
if (glb_is_classification && !glb_is_binomial) {
df[, glb_rsp_var_out] <- predict(glb_fin_mdl, newdata=df, type="raw")
}
return(df)
}
glb_trnent_df <- glb_get_predictions(df=glb_trnent_df)## Warning in glb_get_predictions(df = glb_trnent_df): opt.prob.threshold.fit
## differs for fin_mdl: 0.5 vs. sel_mdl: 0.6print(glb_feats_df <- mymerge_feats_importance(feats_df=glb_feats_df, sel_mdl=glb_fin_mdl,
entity_df=glb_trnent_df))## id cor.y exclude.as.feat cor.y.abs
## 25 NewsDesk.fctr -0.115969991 0 0.115969991
## 52 WordCount 0.257526549 0 0.257526549
## 48 SectionName.fctr -0.089486409 0 0.089486409
## 28 PubDate.hour 0.159167673 0 0.159167673
## 29 PubDate.minute -0.031469083 0 0.031469083
## 31 PubDate.second -0.012253600 0 0.012253600
## 19 H.num.chars.log -0.171062360 0 0.171062360
## 18 H.num.chars -0.147211183 0 0.147211183
## 27 PubDate.date -0.012014100 0 0.012014100
## 37 S.num.chars -0.178433540 0 0.178433540
## 5 A.num.chars -0.176187472 0 0.176187472
## 6 A.num.chars.log -0.205487403 0 0.205487403
## 38 S.num.chars.log -0.205402971 0 0.205402971
## 32 PubDate.wkday -0.039801288 0 0.039801288
## 49 SectionName.fctr -0.089486409 0 0.089486409
## 24 NewsDesk.fctr -0.115969991 0 0.115969991
## 30 PubDate.month 0.019148739 0 0.019148739
## 23 H.num.words.unq.log -0.208052027 0 0.208052027
## 20 H.num.words -0.190083728 0 0.190083728
## 22 H.num.words.unq -0.193599112 0 0.193599112
## 21 H.num.words.log -0.204378687 0 0.204378687
## 41 S.num.words.unq -0.213378613 0 0.213378613
## 42 S.num.words.unq.log -0.244677426 0 0.244677426
## 9 A.num.words.unq -0.211522561 0 0.211522561
## 8 A.num.words.log -0.239527732 0 0.239527732
## 10 A.num.words.unq.log -0.244569301 0 0.244569301
## 40 S.num.words.log -0.239719045 0 0.239719045
## 39 S.num.words -0.207846930 0 0.207846930
## 7 A.num.words -0.205674056 0 0.205674056
## 50 SubsectionName.fctr -0.137898635 0 0.137898635
## 45 S.will -0.060392694 0 0.060392694
## 36 S.new -0.069088180 0 0.069088180
## 4 A.new -0.069393513 0 0.069393513
## 13 A.will -0.060842964 0 0.060842964
## 43 S.time -0.057595102 0 0.057595102
## 11 A.time -0.057790617 0 0.057790617
## 44 S.week -0.084814939 0 0.084814939
## 12 A.week -0.084814939 0 0.084814939
## 46 S.year -0.051235765 0 0.051235765
## 14 A.year -0.051235765 0 0.051235765
## 34 S.compani -0.053012962 0 0.053012962
## 2 A.compani -0.053099633 0 0.053099633
## 17 H.new -0.079839251 0 0.079839251
## 15 A.york -0.064751157 0 0.064751157
## 47 S.york -0.064751157 0 0.064751157
## 1 .rnorm -0.008703337 0 0.008703337
## 3 A.has.http -0.013592603 0 0.013592603
## 16 H.has.http NA 0 NA
## 26 Popular 1.000000000 1 1.000000000
## 33 PubDate.year NA 0 NA
## 35 S.has.http NA 0 NA
## 51 UniqueID 0.011824920 1 0.011824920
## cor.high.X is.ConditionalX.y is.cor.y.abs.low importance
## 25 <NA> TRUE FALSE 100.0000000
## 52 <NA> TRUE FALSE 90.9361320
## 48 <NA> TRUE FALSE 46.2256577
## 28 <NA> TRUE FALSE 41.1964054
## 29 <NA> TRUE FALSE 28.4647014
## 31 <NA> TRUE FALSE 23.2904197
## 19 H.num.chars TRUE FALSE 22.0807744
## 18 <NA> TRUE FALSE 21.9376223
## 27 <NA> TRUE FALSE 21.4302834
## 37 A.num.chars TRUE FALSE 18.1218961
## 5 <NA> TRUE FALSE 16.9166846
## 6 S.num.chars.log TRUE FALSE 16.4786255
## 38 <NA> TRUE FALSE 15.7366530
## 32 <NA> TRUE FALSE 15.0538430
## 49 <NA> TRUE FALSE 8.7006298
## 24 <NA> TRUE FALSE 6.9410469
## 30 <NA> TRUE FALSE 6.9002002
## 23 H.num.chars.log TRUE FALSE 6.3970600
## 20 <NA> TRUE FALSE 6.3225781
## 22 H.num.words TRUE FALSE 6.3166418
## 21 <NA> TRUE FALSE 6.1230165
## 41 S.num.words TRUE FALSE 5.4969604
## 42 S.num.chars TRUE FALSE 5.4344611
## 9 <NA> TRUE FALSE 5.3449758
## 8 <NA> TRUE FALSE 5.2480939
## 10 <NA> TRUE FALSE 5.1970811
## 40 A.num.words.log TRUE FALSE 5.0951985
## 39 A.num.words TRUE FALSE 5.0572216
## 7 <NA> TRUE FALSE 4.9203295
## 50 <NA> TRUE FALSE 4.8308148
## 45 <NA> TRUE FALSE 1.6633510
## 36 <NA> TRUE FALSE 1.6437976
## 4 S.new TRUE FALSE 1.6122236
## 13 S.will TRUE FALSE 1.6036419
## 43 <NA> TRUE FALSE 1.4513837
## 11 S.time TRUE FALSE 1.4330261
## 44 <NA> TRUE FALSE 1.4094539
## 12 S.week TRUE FALSE 1.2975671
## 46 <NA> TRUE FALSE 1.2405571
## 14 S.year TRUE FALSE 1.1671384
## 34 <NA> TRUE FALSE 1.1157192
## 2 S.compani TRUE FALSE 1.0714318
## 17 <NA> TRUE FALSE 1.0122754
## 15 S.york TRUE FALSE 0.4472829
## 47 <NA> TRUE FALSE 0.4339125
## 1 <NA> TRUE FALSE NA
## 3 <NA> FALSE FALSE NA
## 16 <NA> FALSE NA NA
## 26 <NA> NA FALSE NA
## 33 <NA> FALSE NA NA
## 35 <NA> FALSE NA NA
## 51 <NA> NA FALSE NA# Used again in predict.data.new chunk
glb_analytics_diag_plots <- function(obs_df) {
if (length(vars <- subset(glb_feats_df, importance > 0)$id) > 5) {
warning("Limiting important feature scatter plots to 5 out of ", length(vars))
vars <- vars[1:5]
}
for (var in vars) {
plot_df <- melt(obs_df, id.vars=var,
measure.vars=c(glb_rsp_var, glb_rsp_var_out))
# if (var == "<feat_name>") print(myplot_scatter(plot_df, var, "value",
# facet_colcol_name="variable") +
# geom_vline(xintercept=<divider_val>, linetype="dotted")) else
print(myplot_scatter(plot_df, var, "value", colorcol_name="variable",
facet_colcol_name="variable", jitter=TRUE) +
guides(color=FALSE))
}
if (glb_is_regression) {
# plot_vars_df <- subset(glb_feats_df, importance >
# glb_feats_df[glb_feats_df$id == ".rnorm", "importance"])
plot_vars_df <- orderBy(~ -importance, glb_feats_df)
if (nrow(plot_vars_df) == 0)
warning("No important features in glb_fin_mdl") else
print(myplot_prediction_regression(df=obs_df,
feat_x=ifelse(nrow(plot_vars_df) > 1, plot_vars_df$id[2],
".rownames"),
feat_y=plot_vars_df$id[1],
rsp_var=glb_rsp_var, rsp_var_out=glb_rsp_var_out,
id_vars=glb_id_vars)
# + facet_wrap(reformulate(plot_vars_df$id[2])) # if [1 or 2] is a factor
# + geom_point(aes_string(color="<col_name>.fctr")) # to color the plot
)
}
if (glb_is_classification) {
if (nrow(plot_vars_df <- subset(glb_feats_df, importance > 0)) == 0)
warning("No features in selected model are statistically important")
else print(myplot_prediction_classification(df=obs_df,
feat_x=ifelse(nrow(plot_vars_df) > 1, plot_vars_df$id[2],
".rownames"),
feat_y=plot_vars_df$id[1],
rsp_var=glb_rsp_var,
rsp_var_out=glb_rsp_var_out,
id_vars=glb_id_vars)
# + geom_hline(yintercept=<divider_val>, linetype = "dotted")
)
}
}
glb_analytics_diag_plots(obs_df=glb_trnent_df)## Warning in glb_analytics_diag_plots(obs_df = glb_trnent_df): Limiting
## important feature scatter plots to 5 out of 45
## UniqueID NewsDesk SectionName SubsectionName
## 96 96 Business Business Day Dealbook
## 6370 6370
## Headline
## 96 Morning Agenda: Evans Bank Under Fire
## 6370 Latest Updates: Protests Nationwide as More Troops Are Called to Ferguson
## Snippet
## 96 New York set to accuse Evans Bank of redlining. | Dollar General raises bid for Family Dollar. | Public pension funds staying mum on inversions. | Banks pursuing military customers.
## 6370 Following the events the day and evening after the release of the grand jury decision in the shooting of Michael Brown in Ferguson. Mo.
## Abstract
## 96 New York set to accuse Evans Bank of redlining. | Dollar General raises bid for Family Dollar. | Public pension funds staying mum on inversions. | Banks pursuing military customers.
## 6370 Following the events the day and evening after the release of the grand jury decision in the shooting of Michael Brown in Ferguson. Mo.
## WordCount PubDate Popular PubDate.year PubDate.month
## 96 0 2014-09-02 07:57:53 0 2014 9
## 6370 10912 2014-11-25 07:49:49 1 2014 11
## PubDate.date PubDate.wkday PubDate.hour PubDate.minute PubDate.second
## 96 2 2 7 57 53
## 6370 25 2 7 49 49
## NewsDesk.fctr SectionName.fctr SubsectionName.fctr .rnorm
## 96 Business Business Day Dealbook 0.1410843
## 6370 -0.4313549
## Popular.fctr H.new H.has.http H.num.chars H.num.words H.num.words.unq
## 96 N 0 0 37 5 5
## 6370 Y 0 0 73 7 7
## H.num.chars.log H.num.words.log H.num.words.unq.log S.compani S.new
## 96 3.637586 1.791759 1.791759 0 1
## 6370 4.304065 2.079442 2.079442 0 0
## S.time S.week S.will S.year S.york S.has.http S.num.chars S.num.words
## 96 0 0 0 0 1 0 181 23
## 6370 0 0 0 0 0 0 135 12
## S.num.words.unq S.num.chars.log S.num.words.log S.num.words.unq.log
## 96 21 5.204007 3.178054 3.091042
## 6370 12 4.912655 2.564949 2.564949
## A.compani A.new A.time A.week A.will A.year A.york A.has.http
## 96 0 1 0 0 0 0 1 0
## 6370 0 0 0 0 0 0 0 0
## A.num.chars A.num.words A.num.words.unq A.num.chars.log
## 96 181 23 21 5.204007
## 6370 135 12 12 4.912655
## A.num.words.log A.num.words.unq.log
## 96 3.178054 3.091042
## 6370 2.564949 2.564949
## Popular.fctr.predict.Final.rf.prob Popular.fctr.predict.Final.rf
## 96 0.000 N
## 6370 0.746 Y
## Popular.fctr.predict.Final.rf.accurate .label
## 96 TRUE 96
## 6370 TRUE 6370
replay.petrisim(pn=glb_analytics_pn,
replay.trans=(glb_analytics_avl_objs <- c(glb_analytics_avl_objs,
"data.training.all.prediction","model.final")), flip_coord=TRUE)## time trans "bgn " "fit.data.training.all " "predict.data.new " "end "
## 0.0000 multiple enabled transitions: data.training.all data.new model.selected firing: data.training.all
## 1.0000 1 2 1 0 0
## 1.0000 multiple enabled transitions: data.training.all data.new model.selected model.final data.training.all.prediction firing: data.new
## 2.0000 2 1 1 1 0
## 2.0000 multiple enabled transitions: data.training.all data.new model.selected model.final data.training.all.prediction data.new.prediction firing: model.selected
## 3.0000 3 0 2 1 0
## 3.0000 multiple enabled transitions: model.final data.training.all.prediction data.new.prediction firing: data.training.all.prediction
## 4.0000 5 0 1 1 1
## 4.0000 multiple enabled transitions: model.final data.training.all.prediction data.new.prediction firing: model.final
## 5.0000 4 0 0 2 1
glb_script_df <- rbind(glb_script_df,
data.frame(chunk_label="predict.data.new",
chunk_step_major=max(glb_script_df$chunk_step_major)+1,
chunk_step_minor=0,
elapsed=(proc.time() - glb_script_tm)["elapsed"]))
print(tail(glb_script_df, 2))## chunk_label chunk_step_major chunk_step_minor elapsed
## elapsed13 fit.data.training 7 1 539.919
## elapsed14 predict.data.new 8 0 550.030Step 8: predict data.new
# Compute final model predictions
glb_newent_df <- glb_get_predictions(glb_newent_df)## Warning in glb_get_predictions(glb_newent_df): opt.prob.threshold.fit
## differs for fin_mdl: 0.5 vs. sel_mdl: 0.6glb_analytics_diag_plots(obs_df=glb_newent_df)## Warning in glb_analytics_diag_plots(obs_df = glb_newent_df): Limiting
## important feature scatter plots to 5 out of 45## Warning in min(x): no non-missing arguments to min; returning Inf## Warning in max(x): no non-missing arguments to max; returning -Inf## Warning: Removed 1870 rows containing missing values (geom_point).## Warning: Removed 1870 rows containing missing values (geom_point).
## Warning in min(x): no non-missing arguments to min; returning Inf## Warning in max(x): no non-missing arguments to max; returning -Inf## Warning: Removed 1870 rows containing missing values (geom_point).## Warning: Removed 1870 rows containing missing values (geom_point).
## Warning in min(x): no non-missing arguments to min; returning Inf## Warning in max(x): no non-missing arguments to max; returning -Inf## Warning: Removed 1870 rows containing missing values (geom_point).## Warning: Removed 1870 rows containing missing values (geom_point).
## Warning in min(x): no non-missing arguments to min; returning Inf## Warning in max(x): no non-missing arguments to max; returning -Inf## Warning: Removed 1870 rows containing missing values (geom_point).## Warning: Removed 1870 rows containing missing values (geom_point).
## Warning in min(x): no non-missing arguments to min; returning Inf## Warning in max(x): no non-missing arguments to max; returning -Inf## Warning: Removed 1870 rows containing missing values (geom_point).## Warning: Removed 1870 rows containing missing values (geom_point).
## UniqueID NewsDesk SectionName SubsectionName
## 6563 6563
## 6753 6753 Metro N.Y. / Region
## Headline
## 6563 First Draft Focus: A Red Ribbon on the White House
## 6753 Reaction to Eric Garner Grand Jury Decision
## Snippet
## 6563 A large red ribbon was hung from the White House on Monday in honor of World AIDS Day.
## 6753 Reactions to a grand jurys decision not to indict the police officer who caused the death of a Staten Island man by putting him in a chokehold.
## Abstract
## 6563 A large red ribbon was hung from the White House on Monday in honor of World AIDS Day.
## 6753 Reactions to a grand jurys decision not to indict the police officer who caused the death of a Staten Island man by putting him in a chokehold.
## WordCount PubDate Popular PubDate.year PubDate.month
## 6563 0 2014-12-01 15:56:36 NA 2014 12
## 6753 5956 2014-12-03 14:22:22 NA 2014 12
## PubDate.date PubDate.wkday PubDate.hour PubDate.minute PubDate.second
## 6563 1 1 15 56 36
## 6753 3 3 14 22 22
## NewsDesk.fctr SectionName.fctr SubsectionName.fctr .rnorm
## 6563 0.7064699
## 6753 Metro N.Y. / Region 0.7189705
## Popular.fctr H.new H.has.http H.num.chars H.num.words H.num.words.unq
## 6563 <NA> 0 0 50 7 7
## 6753 <NA> 0 0 43 6 6
## H.num.chars.log H.num.words.log H.num.words.unq.log S.compani S.new
## 6563 3.931826 2.079442 2.079442 0 0
## 6753 3.784190 1.945910 1.945910 0 0
## S.time S.week S.will S.year S.york S.has.http S.num.chars S.num.words
## 6563 0 0 0 0 0 0 86 11
## 6753 0 0 0 0 0 0 143 14
## S.num.words.unq S.num.chars.log S.num.words.log S.num.words.unq.log
## 6563 11 4.465908 2.484907 2.484907
## 6753 14 4.969813 2.708050 2.708050
## A.compani A.new A.time A.week A.will A.year A.york A.has.http
## 6563 0 0 0 0 0 0 0 0
## 6753 0 0 0 0 0 0 0 0
## A.num.chars A.num.words A.num.words.unq A.num.chars.log
## 6563 86 11 11 4.465908
## 6753 143 14 14 4.969813
## A.num.words.log A.num.words.unq.log
## 6563 2.484907 2.484907
## 6753 2.708050 2.708050
## Popular.fctr.predict.Final.rf.prob Popular.fctr.predict.Final.rf
## 6563 0.002 N
## 6753 0.460 Y
## Popular.fctr.predict.Final.rf.accurate .label
## 6563 NA 6563
## 6753 NA 6753## Warning: Removed 1870 rows containing missing values (geom_point).
submit_df <- glb_newent_df[, c(glb_id_vars, paste0(glb_rsp_var_out, ".prob"))]
names(submit_df)[2] <- "Probability1"
write.csv(submit_df, paste0(glb_out_pfx, "submit.csv"), row.names=FALSE)
print(orderBy(~ -max.auc.OOB, glb_models_df[, c("model_id",
"max.auc.OOB", "max.Accuracy.OOB")]))## model_id max.auc.OOB max.Accuracy.OOB
## 9 Conditional.X.glm 0.9327051 0.8974234
## 8 Low.cor.X.glm 0.9326067 0.8993680
## 12 Conditional.X.rf 0.9287358 0.9071463
## 13 Conditional.X.no.rnorm.rf 0.9283871 0.9076325
## 11 Conditional.X.cp.0.rpart 0.8910062 0.9027710
## 10 Conditional.X.rpart 0.8225030 0.8930481
## 7 Interact.High.cor.Y.glm 0.8156980 0.8142927
## 6 Max.cor.Y.glm 0.7378291 0.7681089
## 4 Max.cor.Y.cv.0.cp.0.rpart 0.6552144 0.7438017
## 1 MFO.myMFO_classfr 0.5000000 0.8327662
## 3 Max.cor.Y.cv.0.rpart 0.5000000 0.8327662
## 5 Max.cor.Y.rpart 0.5000000 0.8327662
## 2 Random.myrandom_classfr 0.4821958 0.1672338
## 14 Final.rf NA NAprint(sprintf("glb_sel_mdl_id: %s", glb_sel_mdl_id))## [1] "glb_sel_mdl_id: Conditional.X.no.rnorm.rf"print(sprintf("glb_fin_mdl_id: %s", glb_fin_mdl_id))## [1] "glb_fin_mdl_id: Final.rf"print(dim(glb_fitent_df))## [1] 4475 58# tmp_replay_lst <- replay.petrisim(pn=glb_analytics_pn,
# replay.trans=(glb_analytics_avl_objs <- c(glb_analytics_avl_objs,
# "data.new.prediction")), flip_coord=TRUE)
# print(ggplot.petrinet(tmp_replay_lst[["pn"]]) + coord_flip())Null Hypothesis (\(\sf{H_{0}}\)): mpg is not impacted by am_fctr.
The variance by am_fctr appears to be independent. #{r q1, cache=FALSE} # print(t.test(subset(cars_df, am_fctr == "automatic")$mpg, # subset(cars_df, am_fctr == "manual")$mpg, # var.equal=FALSE)$conf) # We reject the null hypothesis i.e. we have evidence to conclude that am_fctr impacts mpg (95% confidence). Manual transmission is better for miles per gallon versus automatic transmission.
## chunk_label chunk_step_major chunk_step_minor elapsed
## 12 fit.models 6 2 433.449
## 14 fit.data.training 7 1 539.919
## 7 select.features 4 0 85.848
## 11 fit.models 6 1 180.884
## 9 partition.data.training 5 0 124.117
## 10 fit.models 6 0 140.736
## 13 fit.data.training 7 0 447.799
## 15 predict.data.new 8 0 550.030
## 6 extract.features 3 0 5.551
## 8 remove.correlated.features 4 1 88.177
## 4 manage_missing_data 2 2 2.831
## 2 cleanse_data 2 0 0.864
## 5 encodeORretype.data 2 3 3.123
## 3 inspectORexplore.data 2 1 0.903
## 1 import_data 1 0 0.002
## elapsed_diff
## 12 252.565
## 14 92.120
## 7 80.297
## 11 40.148
## 9 35.940
## 10 16.619
## 13 14.350
## 15 10.111
## 6 2.428
## 8 2.329
## 4 1.928
## 2 0.862
## 5 0.292
## 3 0.039
## 1 0.000## [1] "Total Elapsed Time: 550.03 secs"
## R version 3.1.3 (2015-03-09)
## Platform: x86_64-apple-darwin13.4.0 (64-bit)
## Running under: OS X 10.10.3 (Yosemite)
##
## locale:
## [1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8
##
## attached base packages:
## [1] tcltk grid stats graphics grDevices utils datasets
## [8] methods base
##
## other attached packages:
## [1] randomForest_4.6-10 rpart.plot_1.5.2 rpart_4.1-9
## [4] ROCR_1.0-7 gplots_2.16.0 caTools_1.17.1
## [7] caret_6.0-41 lattice_0.20-31 tm_0.6
## [10] NLP_0.1-6 sqldf_0.4-10 RSQLite_1.0.0
## [13] DBI_0.3.1 gsubfn_0.6-6 proto_0.3-10
## [16] reshape2_1.4.1 plyr_1.8.1 doBy_4.5-13
## [19] survival_2.38-1 ggplot2_1.0.1
##
## loaded via a namespace (and not attached):
## [1] bitops_1.0-6 BradleyTerry2_1.0-6 brglm_0.5-9
## [4] car_2.0-25 chron_2.3-45 class_7.3-12
## [7] codetools_0.2-11 colorspace_1.2-6 compiler_3.1.3
## [10] digest_0.6.8 e1071_1.6-4 evaluate_0.5.5
## [13] foreach_1.4.2 formatR_1.1 gdata_2.13.3
## [16] gtable_0.1.2 gtools_3.4.1 htmltools_0.2.6
## [19] iterators_1.0.7 KernSmooth_2.23-14 knitr_1.9
## [22] labeling_0.3 lme4_1.1-7 MASS_7.3-40
## [25] Matrix_1.2-0 mgcv_1.8-6 minqa_1.2.4
## [28] munsell_0.4.2 nlme_3.1-120 nloptr_1.0.4
## [31] nnet_7.3-9 parallel_3.1.3 pbkrtest_0.4-2
## [34] quantreg_5.11 RColorBrewer_1.1-2 Rcpp_0.11.5
## [37] rmarkdown_0.5.1 scales_0.2.4 slam_0.1-32
## [40] SparseM_1.6 splines_3.1.3 stringr_0.6.2
## [43] tools_3.1.3 yaml_2.1.13